Formation of Molecular Clouds and Global Conditions for Star
Formation
Clare L. Dobbs
University of Exeter
Mark R. Krumholz
arXiv:1312.3223v1 [astro-ph.GA] 11 Dec 2013
University of California, Santa Cruz
Javier Ballesteros-Paredes
Universidad Nacional Autónoma de México
Alberto D. Bolatto
University of Maryland, College Park
Yasuo Fukui
Nagoya University
Mark Heyer
University of Massachusetts
Mordecai-Mark Mac Low
American Museum of Natural History
Eve C. Ostriker
Princeton University
Enrique Vázquez-Semadeni
Universidad Nacional Autónoma de México
Giant molecular clouds (GMCs) are the primary reservoirs of cold, star-forming molecular
gas in the Milky Way and similar galaxies, and thus any understanding of star formation
must encompass a model for GMC formation, evolution, and destruction. These models are
necessarily constrained by measurements of interstellar molecular and atomic gas, and the
emergent, newborn stars. Both observations and theory have undergone great advances in recent
years, the latter driven largely by improved numerical simulations, and the former by the advent
of large-scale surveys with new telescopes and instruments. This chapter offers a thorough
review of the current state of the field.
1. INTRODUCTION
rent status of observations and theory of molecular clouds,
focusing on key advances in the field since Protostars and
Planets V.
The first detections of molecules in the ISM date from
the 1930s, with the discovery of CH and CN within the
diffuse interstellar bands (Swings and Rosenfeld 1937;
McKellar 1940) and later the microwave lines of OH
(Weinreb et al. 1963), NH3 (Cheung et al. 1968), water vapor (Cheung et al. 1969) and H2 CO (Snyder et al. 1969).
Progress accelerated in the 1970s with the first measurements of molecular hydrogen (Carruthers 1970) and the
12
CO J=1-0 line at 2.6mm (Wilson et al. 1970) and the con-
Stars form in a cold, dense, molecular phase of the interstellar medium (ISM) that appears to be organized into
coherent, localized volumes or clouds. The star formation
history of the universe, the evolution of galaxies, and the
formation of planets in stellar environments are all coupled
to the formation of these clouds, the collapse of unstable
regions within them to stars, and the clouds’ final dissipation. The physics of these regions is complex, and descriptions of cloud structure and evolution remain incomplete
and require continued exploration. Here we review the cur1
tinued development of millimeter wave instrumentation and the same time, Roberts (1969) showed that the gas refacilities.
sponse to a stellar spiral arm produces a strong spiral
The first maps of CO emission in nearby star forming shock, likely observed as dust lanes and associated with
regions and along the Galactic Plane revealed the unex- molecular gas. Somewhat more recently, the idea of cloud
pectedly large spatial extent of giant molecular clouds formation from turbulent flows in the ISM has emerged
(GMCs Kutner et al. 1977; Lada 1976; Blair et al. 1978; (Ballesteros-Paredes et al. 1999), as well as colliding flows
Blitz and Thaddeus 1980), and their substantial contri- from stellar feedback processes (Koyama and Inutsuka
bution to the mass budget of the ISM (Scoville 1975; 2000).
Gordon and Burton 1976; Burton and Gordon 1978; Sanders et al.The nature of GMCs, their lifetime, and whether they are
1984). Panoramic imaging of 12 CO emission in the virialized, remains unclear. Early models of cloud-cloud
Milky Way from both the Northern and Southern Hemi- collisions required very long-lasting clouds (100 Myr) in
spheres enabled the first complete view of the molecular order to build up more massive GMCs (Kwan 1979). Since
gas distribution in the Galaxy (Dame et al. 1987, 2001) then, lifetimes have generally been revised downwards.
and the compilation of GMC properties (Solomon et al. Several observationally derived estimates, including up to
1987; Scoville et al. 1987). Higher angular resolution ob- the present date, have placed cloud lifetimes at around 20–
servations of optically thin tracers of molecular gas in 30 Myr (Bash et al. 1977; Blitz and Shu 1980; Fukui et al.
nearby clouds revealed a complex network of filaments 1999; Kawamura et al. 2009; Miura et al. 2012), although
(Bally et al. 1987; Heyer et al. 1987), and high density trac- there have been longer estimates for molecule rich galaxies
ers such as NH3 , CS, and HCN revealed the dense regions (Tomisaka 1986; Koda et al. 2009) and shorter estimates for
of active star formation (Myers 1983; Snell et al. 1984). smaller, nearby clouds (Elmegreen 2000; Hartmann et al.
Since this early work, large, millimeter filled aperture 2001).
(IRAM 30m, NRO 45m), interferometric (BIMA, OVRO,
In the 1980s and 1990s, GMCs were generally thought
Plateau de Bure) and submillimeter (CSO, JCMT) facil- to be supported against gravitational collapse, and in virial
ities have provided improved sensitivity and the ability equilibrium. Magnetic fields were generally favored as a
to measure higher excitation conditions. Observations to means of support (Shu et al. 1987). Turbulence would dissidate have identified ∼200 distinct interstellar molecules pate unless replenished, whilst rotational support was found
(van Dishoeck and Blake 1998; Müller et al. 2005), and the to be insufficient (e.g., Silk 1980). More recently these conlast 40 years of observations using these molecules have clusions have been challenged by new observations, which
determined a set of cloud properties on which our limited have revised estimates of magnetic field strengths downunderstanding of cloud physics is based.
wards, and new simulations and theoretical models that sugTheoretically, the presence of molecular hydrogen in the gest that clouds may in fact be turbulence-supported, or that
ISM was predicted long before the development of large they may be entirely transient objects that are not supported
scale CO surveys (e.g., Spitzer 1949). In the absence of against collapse at all. These questions are all under active
metals, formation of H2 by gas phase reactions catalyzed discussion, as we review below.
by electrons and protons is extremely slow, but dust grains
In § 2, we describe the main new observational results,
catalyze the reaction and speed it up by orders of magni- and corresponding theoretical interpretations. These intude. As a result, H2 formation is governed by the density clude the extension of the Schmidt-Kennicutt relation to
of dust grains, gas density, and the ability of hydrogen other tracers, notably H2 , as well as to much smaller scales,
atoms to stick to dust grains and recombine (van de Hulst e.g., those of individual clouds. § 2 also examines the latest
1948; McCrea and McNally 1960; Gould and Salpeter results on GMC properties, both within the Milky Way and
1963; Hollenbach and Salpeter 1971). The ISM exhibits in external galaxies. Compared to the data that were availa sharp transition in molecular fraction from low to high able at the time of PPV, CO surveys offer much higher resdensities, typically at 1–100 cm−3 (or Σ ∼1–100 M⊙ olution and sensitivity within the Milky Way, and are able
pc−2 ), dependent mostly on the UV radiation field and to better cover a wider range of environments beyond it. In
metallicity (van Dishoeck and Black 1986; Pelupessy et al. § 3, we discuss GMC formation, providing a summary of
2006; Glover and Mac Low 2007a; Dobbs et al. 2008; the main background and theory, whilst reporting the main
Krumholz et al. 2008, 2009a; Gnedin et al. 2009). This advances in numerical simulations since PPV. We also disdramatic increase in H2 fraction represents a change to the cuss progress on calculating the conversion of atomic to
regime where H2 becomes self shielding. Many processes molecular gas, and CO chemistry. § 4 describes the various
have been invoked to explain how atomic gas reaches the scenarios for the evolution of GMCs, including the revival
densities (& 100 cm−3 ) required to become predominantly of globally collapsing clouds as a viable theoretical model,
molecular (see Section 3). Several mechanisms likely and examines the role of different forms of stellar feedback
to govern ISM structure became apparent in the 1960s: as internal and external sources of cloud motions. Then in
cloud-cloud collisions (Oort 1954; Field and Saslaw 1965), § 5 we relate the star forming properties in GMCs to these
gravitational instabilities (e.g., Goldreich and Lynden-Bell different scenarios. Finally in § 6 we look forward to what
1965a), thermal instabilities (Field 1965), and magnetic we can expect between now and PPVII.
instabilities (Parker 1966; Mouschovias 1974). At about
2
2. OBSERVED PROPERTIES OF GMCs
to-H2 conversion factor that depends on the local dust-togas ratio removes most of the variation in τdep (H2 ).
On scales of a few hundred parsecs, the scatter in
τdep (H2 ) rises significantly (e.g., Schruba et al. 2010;
Onodera et al. 2010) and the SFR–H2 correlation breaks
down. This is partially a manifestation of the large dispersion in SFR per unit mass in individual GMCs (Lada et al.
2010), but it is also a consequence of the time scales involved (Kawamura et al. 2009; Kim et al. 2013). Technical
issues concerning the interpretation of the tracers also become important on the small scales (Calzetti et al. 2012).
On sub-GMC scales there are strong correlations between star formation and extinction, column density, and
volume density. The correlation with volume density is
very close to that observed in ultraluminous IR galaxies
(Wu et al. 2005). Some authors have interpreted these
data as implying that star formation only begins above
a threshold column density of ΣH2 ∼110–130 M⊙ yr−1
or volume density n ∼ 104−5 cm−3 (Evans et al. 2009;
Heiderman et al. 2010; Lada et al. 2010, 2012). However, others argue that the data are equally consistent with
a smooth rise in SFR with volume or surface density,
without any particular threshold value (Krumholz and Tan
2007; Narayanan et al. 2008b,a; Gutermuth et al. 2011;
Krumholz et al. 2012; Burkert and Hartmann 2013).
2.1. GMCs and Star Formation
Molecular gas is strongly correlated with star formation on scales from entire galaxies (Kennicutt 1989, 1998;
Gao and Solomon 2004; Saintonge et al. 2011a) to kpc
and sub-kpc regions (Wong and Blitz 2002; Bigiel et al.
2008; Rahman et al. 2012; Leroy et al. 2013) to individual GMCs (Evans et al. 2009; Heiderman et al. 2010;
Lada et al. 2010, 2012). These relations take on different
shapes at different scales. Early studies of whole galaxies found a power-law correlation between total gas content (H I plus H2 ) and star formation rate (SFR) with an
index N ∼ 1.5 (Kennicutt 1989, 1998). These studies include galaxies that span a very large range of properties,
from dwarfs to ultraluminous IR galaxies, so it is possible that the physical underpinnings of this relation are
different in different regimes. Transitions with higher critical densities such as HCN(1 − 0) and higher-J CO lines
(Gao and Solomon 2004; Bayet et al. 2009; Juneau et al.
2009; Garcı́a-Burillo et al. 2012) also show power-law correlations but with smaller indices; the index appears to depend mostly on the line critical density, a result that can be
explained through models (Krumholz and Thompson 2007;
Narayanan et al. 2008b,a).
Within galaxies the star formation rate surface density, ΣSF R , is strongly correlated with the surface density of molecular gas as traced by CO emission and
only very weakly, if at all, related to atomic gas. The
strong correlation with H2 persists even in regions where
atomic gas dominates the mass budget (Schruba et al. 2011;
Bolatto et al. 2011). The precise form of the SFR–H2 correlation is a subject of study, with results spanning the range
from super-linear (Kennicutt et al. 2007; Liu et al. 2011;
Calzetti et al. 2012) to approximately linear (Bigiel et al.
2008; Blanc et al. 2009; Rahman et al. 2012; Leroy et al.
2013) to sub-linear (Shetty et al. 2013). Because CO is
used to trace H2 , the correlation can be altered by systematic variations in the CO to H2 conversion factor, an effect
that may flatten the observed relation compared to the true
one (Shetty et al. 2011; Narayanan et al. 2011, 2012).
The SFR–H2 correlation defines a molecular depletion
time, τdep (H2 ) = M (H2 )/SFR, which is the time required
to consume all the H2 at the current SFR. A linear SFR–
H2 correlation implies a constant τdep (H2 ), while superlinear (sub-linear) relations yield a time scale τdep (H2 )
that decreases (increases) with surface density. In regions
where CO emission is present, the mean depletion time over
kpc scales is τdep (H2 ) = 2.2 Gyr with ±0.3 dex scatter,
with some dependence on the local conditions (Leroy et al.
2013). Saintonge et al. (2011) find that, for entire galaxies,
τdep (H2 ) decreases by a factor of ∼ 3 over two orders of
magnitude increase in the SFR surface. Leroy et al. (2013)
show that the kpc-scale measurements within galaxies are
consistent with this trend, but that τdep (H2 ) also correlates
with the dust-to-gas ratio. For normal galaxies, using a CO-
2.2. GMCs as a Component of the Interstellar Medium
Molecular clouds are the densest, coldest, highest column density, highest extinction component of the interstellar medium. Their masses are dominated by molecular gas
(H2 ), with a secondary contribution from He (∼ 26%),
and a varying contribution from H I in a cold envelope
(e.g., Fukui and Kawamura 2010) and interclump gas detectable by H I self-absorption (Goldsmith and Li 2005).
Most of the molecular mass in galaxies is in the form of
molecular clouds, with the possible exception of galaxies with gas surface densities substantially higher than that
of the Milky Way, where a substantial diffuse H2 component exists (Papadopoulos et al. 2012b,a; Pety et al. 2013;
Colombo et al. 2013).
Molecular cloud masses range from ∼ 102 M⊙ for
small clouds at high Galactic latitudes (e.g., Magnani et al.
1985) and in the outer disk of the Milky Way (e.g.,
Brand and Wouterloot 1995; Heyer et al. 2001) up to giant ∼ 107 M⊙ clouds in the central molecular zone of the
Galaxy (Oka et al. 2001). The measured mass spectrum
of GMCs (see §2.3) implies that most of the molecular
mass resides in the largest GMCs. Bulk densities of clouds
are log[nH2 /cm−3 ] = 2.6 ± 0.3 (Solomon et al. 1987;
Roman Duval et al. 2010), but clouds have inhomogenous density distributions with large contrasts (Stutzki et al.
1988). The ratio of molecular to stellar mass in galaxies
shows a strong trend with galaxy color from high in blue
galaxies (10% for N U V − r ∼ 2) to low in red galaxies
(. 0.16% for N U V − r & 5) (Saintonge et al. 2011a).
The typical molecular to atomic ratio in galaxies where
3
both H I and H2 are detected is Rmol ≡ MH2 /MHI ≈ 0.3
with scatter of ±0.4 dex. The large scatter reflects the fact
that the atomic and molecular masses are only weakly correlated, and in contrast with the molecular gas to stellar
mass fraction, the ratio Rmol shows only weak correlations
with galaxy properties such as color (Leroy et al. 2005;
Saintonge et al. 2011a).
In terms of their respective spatial distributions, in spiral galaxies H2 is reasonably well described by an exponential profile with a scale length ℓCO ≈ 0.2 R25 ,
rather smaller than the optical emission (Young et al. 1995;
Regan et al. 2001; Leroy et al. 2009; Schruba et al. 2011),
where R25 is the 25th magnitude isophotal radius for
the stellar light distribution. In contrast, H I shows a
nearly flat distribution with typical maximum surface density ΣHI,max ∼ 12 M⊙ pc−2 (similar to the H I column
seen toward Solar neighborhood clouds, Lee et al. 2012).
Galaxy centers are the regions that show the most variability and the largest departures from these trends (Regan et al.
2001; Bigiel and Blitz 2012). At low metallicities the H I
surface density can be much larger, possibly scaling as
ΣHI,max ∼ Z −1 (Fumagalli et al. 2010; Bolatto et al.
2011; Wong et al. 2013). In spiral galaxies the transition between the atomic- and molecular-dominated regions occurs
at R ∼ 0.4 R25 (e.g., Leroy et al. 2008). The CO emission
also shows much more structure than the H I on the small
scales (Leroy et al. 2013). In spirals with well defined arms
(NGC 6946, M 51, NGC628) the interarm regions contain
at least 30% of the measured CO luminosity (Foyle et al.
2010), but at fixed total gas surface density Rmol is very
similar for arm and interarm regions, suggesting that arms
act mostly to collect gas rather to directly trigger H2 formation (Foyle et al. 2010) (see Figure 1). We discuss the
relationship between H I and H2 in more detail in § 3.3.
olds are required to separate a feature in velocity-crowded
regions. High resolution can also complicate the accounting, as the algorithms may identify cloud substructure as
distinct clouds. Moreover, even once a cloud is identified,
deriving masses and mass-related quantities from observed
CO emission generally requires application of the CO-toH2 conversion factor or the H2 to 13 CO abundance ratio,
both of which can vary within and between clouds in response to local conditions of UV irradiance, density, temperature, and metallicity (Bolatto et al. 2013; Ripple et al.
2013). Millimeter wave interferometers can resolve large
GMC complexes in nearby galaxies but must also account
for missing flux from an extended component of emission.
Despite these observational difficulties, there are some
robust results. Over the mass range M > 104 M⊙ where it
can be measured reliably, the cloud mass spectrum is wellfit by a powerlaw dN/dM ∼ M −γ (cumulative distribution function N (> M ) ∼ M −γ+1 ), with values γ < 2
indicating that most of the mass is in large clouds. For
GMCs in the Milky Way, γ is consistently found to be
in the range 1.5 to 1.8 (Solomon et al. 1987; Kramer et al.
1998; Heyer et al. 2001; Roman Duval et al. 2010) with the
higher value likely biased by the inclusion of cloud fragments identified as distinct clouds. GMCs in the Magellanic
Clouds exhibit a steeper mass function overall and specifically for massive clouds (Fukui et al. 2008; Wong et al.
2011). In M33, γ ranges from 1.6 in the inner regions to 2.3
at larger radii (Rosolowsky and Blitz 2005; Gratier et al.
2012).
In addition to clouds’ masses, we can measure their sizes
and thus their surface densities. The Solomon et al. (1987)
catalog of inner Milky Way GMCs, updated to the current Galactic distance scale, shows a distribution of GMCs
−2
surface densities ΣGMC ≈ 150+95
(±1σ inter−70 M⊙ pc
val) assuming a fixed CO-to-H2 conversion factor XCO =
2 × 1020 cm−2 (K km s−1 )−1 , and including the He mass
(Bolatto et al. 2013). Heyer et al. (2009) re-observed these
clouds in 13 CO and found ΣGMC ∼ 40 M⊙ pc−2 over
the same cloud areas, but concluded that this is likely
at least a factor of 2 too low due to non-LTE and optical depth effects. Heiderman et al. (2010) find that 13 CO
can lead to a factor of 5 underestimate. A reanalysis by
Roman Duval et al. (2010) shows ΣGMC ∼ 144 M⊙ pc−2
using the 13 CO rather than the 12 CO contour to define the
area. Measurements of surface densities in extragalactic
GMCs remain challenging, but with the advent of ALMA
the field is likely to evolve quickly. For a sample of nearby
galaxies, many of them dwarfs, Bolatto et al. (2008) find
ΣGMC ≈ 85 M⊙ pc−2 . Other recent extragalactic surveys find roughly comparable results, ΣGMC ∼ 40 −
170 M⊙ pc−2 (Rebolledo et al. 2012; Donovan Meyer et al.
2013).
GMC surface densities may prove to be a function of
environment. The PAWS survey of M 51 finds a progression in surface density (Colombo et al. 2013), from
clouds in the center (ΣGMC ∼ 210 M⊙ pc−2 ), to clouds
in arms (ΣGMC ∼ 185 M⊙ pc−2 ), to those in interarm
2.3. Statistical Properties of GMCs
Statistical descriptions of GMC properties have provided
insight into the processes that govern their formation and
evolution since large surveys first became possible in the
1980s (see §1). While contemporary observations are more
sensitive and feature better angular resolution and sampling than earlier surveys, identification of clouds within
position-position-velocity (PPV) data cubes remains a significant problem. In practice, one defines a cloud as a set
of contiguous voxels in a PPV data cube of CO emission above a surface brightness threshold. Once a cloud
is defined, one can compute global properties such as size,
velocity dispersion, and luminosity (Williams et al. 1994;
Rosolowsky and Leroy 2006). While these algorithms have
been widely applied, their reliability and completeness
are difficult to evaluate (Ballesteros-Paredes and Mac Low
2002; Pineda et al. 2009; Kainulainen et al. 2009b), particularly for surveys of 12 CO and 13 CO in the Galactic Plane
that are subject to blending of emission from unrelated
clouds. The improved resolution of modern surveys helps
reduce these problems, but higher surface brightness thresh4
Fig. 1.— (left) CO J=1-0 image of M51 from Koda et al. (2009) showing the largest cloud complexes are distributed in spiral arms,
while smaller GMCs lie both in and between spiral features. (right) 3 color image of CO J=1-0 emission from the Taurus molecular
cloud from Narayanan et al. (2008c) illustrating complex gas motions within clouds. Colors represents the CO integrated intensities
over VLSR intervals 0-5 (blue), 5-7.5 (green) and 7.5-12 (red) km s−1 .
regions (ΣGMC ∼ 140 M⊙ pc−2 ). Fukui et al. (2008),
Bolatto et al. (2008), and Hughes et al. (2010) find that
GMCs in the Magellanic Clouds have lower surface densities than those in the inner Milky Way (ΣGMC ∼
50 M⊙ pc−2 ). Because of the presence of extended H2
envelopes at low metallicities (§2.6), however, this may
underestimate their true molecular surface density (e.g.,
Leroy et al. 2009). Even more extreme variations in ΣGMC
are observed near the Galactic Center and in more extreme
starburst environments (see § 2.7).
In addition to studying the mean surface density of
GMCs, observations within the Galaxy can also probe
the distribution of surface densities within GMCs. For a
sample of Solar neighborhood clouds, Kainulainen et al.
(2009a) use infrared extinction measurements to determine that PDFs of column densities are lognormal from
0.5 < AV < 5 (roughly 10–100 M⊙ pc−2 ), with a powerlaw tail at high column densities in actively star-forming
clouds. Column density images derived from dust emission also find such excursions (Schneider et al. 2012, 2013).
Lombardi et al. (2010), also using infrared extinction techniques, find that, although GMCs contain a wide range
of column densities, the mass M and area A contained
within a specified extinction threshold nevertheless obey
the Larson (1981) M ∝ A relation, which implies constant
column density.
Finally, we warn that all column density measurements
are subject to a potential bias. GMCs are identified as contiguous areas with surface brightness values or extinctions
above a threshold typically set by the sensitivity of the data.
Therefore, pixels at or just above this threshold comprise
most of the area of the defined cloud and the measured
cloud surface density is likely biased towards the column
density associated with this threshold limit. Note that there
is also a statistical difference between “mass-weighed” and
“area-weighed” ΣGMC . The former is the average surface
density that contributes most of the mass, while the latter
represents a typical surface density over most of the cloud
extent. Area-weighed ΣGMC tend to be lower, and although
perhaps less interesting from the viewpoint of star formation, they are also easier to obtain from observations.
In addition to mass and area, velocity dispersion is the
third quantity that we can measure for a large sample of
clouds. It provides a coarse assessment of the complex motions in GMCs as illustrated in Figure 1. Larson (1981)
identified scaling relationships between velocity dispersion
and cloud size suggestive of a turbulent velocity spectrum,
and a constant surface density for clouds. Using more
sensitive surveys of GMCs, Heyer et al. (2009) found a
scaling relation that extends the Larson relationships such
that the one-dimensional velocity dispersion σv depends on
the physical radius, R, and the column density ΣGMC , as
5
tigate the scales on which turbulence in molecular clouds
could be driven. They found no break in the velocity
dispersion-size relation, and reported that the second principle component has a “dipole-like” structure. Both features
suggest that the dominant processes driving GMC velocity structure must operate on scales comparable to or larger
than single clouds.
shown in Figure 2. The points follow the expression,
σv = 0.7(ΣGMC /100M⊙ pc−2 )1/2 (R/1 pc)1/2 km s−1 .
(1)
More recent compilations of GMCs in the Milky Way
2.4. Dimensionless Numbers: Virial Parameter and
Mass to Flux Ratio
The virial theorem describes the large-scale dynamics of
gas in GMCs, so ratios of the various terms that appear in it
are a useful guide to what forces are important in GMC evolution. Two of these ratios are the virial parameter, which
evaluates the importance of internal pressure and bulk motion relative to gravity, and the dimensionless mass to flux
ratio, which describes the importance of magnetic fields
compared to gravity. Note, however, that neither of these ratios accounts for potentially-important surface terms (e.g.,
Ballesteros-Paredes et al. 1999).
The virial parameter is defined as αG = Mvirial /MGMC ,
where Mvirial = 5σv2 R/G and MGMC is the luminous
mass of the cloud. For a cloud of uniform density with
negligible surface pressure and magnetic support, αG =
1 corresponds to virial equilibrium and αG = 2 to being marginally gravitationally bound, although in reality
αG > 1 does not strictly imply expansion, nor does αG < 1
strictly imply contraction (Ballesteros-Paredes 2006). Surveys of the Galactic Plane and nearby galaxies using 12 CO
emission to identify clouds find an excellent, near-linear
correlation between Mvirial and the CO luminosity, LCO ,
with a coefficient implying that (for reasonable CO-to-H2
conversion factors) the typical cloud virial parameter is
unity (Solomon et al. 1987; Fukui et al. 2008; Bolatto et al.
2008; Wong et al. 2011). Virial parameters for clouds exhibit a range of values from αG ∼ 0.1 to αG ∼ 10, but
typically αG is indeed ∼ 1. Heyer et al. (2009) reanalyzed the Solomon et al. (1987) GMC sample using 13 CO
J=1-0 emission to derive cloud mass and found a median
αG = 1.9. This value is still consistent with a median
αG = 1, since excitation and abundance variations in
the survey lead to systematic underestimates of MGMC .
A cloud catalog generated directly from the 13 CO emission of the BU-FCRAO Galactic Ring Survey resulted in
a median αG = 0.5 (Roman Duval et al. 2010). Previous surveys (Dobashi et al. 1996; Yonekura et al. 1997;
Heyer et al. 2001) tended to find higher αG for low mass
clouds, possibly a consequence of earlier cloud-finding
algorithms preferentially decomposing single GMCs into
smaller fragments (Bertoldi and McKee 1992).
The importance of magnetic forces is characterized by
the ratio MGMC /Mcr , where Mcr = Φ/(4π 2 G)1/2 and Φ is
the magnetic flux threading the cloud (Mouschovias and Spitzer
1976; Nakano 1978). If MGMC /Mcr > 1 (the supercritical case) then the magnetic field is incapable of providing the requisite force to balance self-gravity, while if
Fig. 2.— The variation of σv /R1/2 with surface density, ΣGMC ,
for Milky Way GMCs from Heyer et al. (2009) (open circles) and
massive cores from Gibson et al. (2009) (blue points). For clarity,
a limited number of error bars are displayed for the GMCs. The
horizontal error bars for the GMCs convey lower limits to the mass
surface density derived from 13 CO. The vertical error bars for both
data sets reflect a 20% uncertainty in the kinematic distances. The
horizontal error bars for the massive cores assume a 50% error in
the C18 O and N2 H+ abundances used to derive mass. The solid
and dotted black lines show loci corresponding to gravitationally
bound and marginally bound clouds respectively. Lines of constant turbulent pressure are illustrated by the red dashed lines. The
mean thermal pressure of the local ISM is shown as the red solid
line.
(Roman Duval et al. 2010) have confirmed this result,
and studies of Local Group galaxies (Bolatto et al. 2008;
Wong et al. 2011) have shown that it applies to GMCs outside the Milky Way as well. Equation 1 is a natural consequence of gravity playing an important role in setting
the characteristic velocity scale in clouds, either through
collapse (Ballesteros-Paredes et al. 2011b) or virial equilibrium (Heyer √
et al. 2009). Unfortunately one expects
only factor of 2 differences in velocity dispersion between clouds that are in free-fall collapse or in virial equilibrium (Ballesteros-Paredes et al. 2011b) making it extremely difficult to distinguish between these possibilities
using observed scaling relations. Concerning the possibility of pressure-confined but mildly self-gravitating
clouds (Field et al. 2011), Figure 2 shows that the turbulent pressures, P = ρσv2 , in observed GMCs are generally
larger than the mean thermal pressure of the diffuse ISM
(Jenkins and Tripp 2011) so these structures must be confined by self-gravity.
As with column density, observations within the Galaxy
can also probe internal velocity structure. Brunt (2003),
Heyer and Brunt (2004), and Brunt et al. (2009) used principal components analysis of GMC velocity fields to inves6
MGMC /Mcr < 1 (the subcritical case) the cloud can be
supported against self-gravity by the magnetic field. Initially subcritical volumes can become supercritical through
ambipolar diffusion (Mouschovias 1987; Lizano and Shu
1989). Evaluating whether a cloud is sub- or supercritical
is challenging. Zeeman measurements of the OH and CN
lines offer a direct measurement of the line of sight component of the magnetic field at densities ∼ 103 and ∼ 105
cm−3 , respectively, but statistical corrections are required
to account for projection effects for both the field and the
column density distribution. Crutcher (2012) provides a
review of techniques and observational results, and report a
mean value MGMC /Mcr ≈ 2 − 3, implying that clouds are
generally supercritical, though not by a large margin.
types indicate the relative lifetimes of each stage, and agedating the star clusters found in type III clouds then makes
it possible to assign absolute durations of 6, 13, and 7 Myrs
for Types I, II, and III respectively. Thus the cumulative
GMC lifetime is τlife ∼ 25 Myrs. This is still substantially
greater than τff , but by less so than in M51.
While lifetime estimates in external galaxies are possible only for large clouds, in the Solar Neighborhood it is
possible to study much smaller clouds, and to do so using timescales derived from the positions of individual stars
on the HR diagram. Elmegreen (2000); Hartmann et al.
(2001) and Ballesteros-Paredes and Hartmann (2007), examining a sample of Solar Neighborhood GMCs, note that
their HR diagrams are generally devoid of post T-Tauri stars
with ages of ∼ 10 Myr or more, suggesting this as an upper limit on τlife . More detailed analysis of HR diagrams,
or other techniques for age-dating stars, generally points
to age spreads of at most ∼ 3 Myr (Reggiani et al. 2011;
Jeffries et al. 2011).
While the short lifetimes inferred for Galactic clouds
might at first seem inconsistent with the extragalactic data,
it is important to remember that the two data sets are probing essentially non-overlapping ranges of cloud mass and
length scale. The largest Solar Neighborhood clouds that
have been age-dated via HR diagrams have masses < 104
M⊙ (the entire Orion cloud is more massive than this, but
the age spreads reported in the literature are only for the few
thousand M⊙ central cluster), below the detection threshold of most extragalactic surveys. Since larger clouds have,
on average, lower densities and longer free-fall timescales,
the difference in τlife is much larger than the difference
in τlife /τff . Indeed, some authors argue that τlife /τff may
be ∼ 10 for Galactic clouds as well as extragalactic ones
(Tan et al. 2006).
2.5. GMC Lifetimes
The natural time unit for GMCs is the free-fall time,
which for a medium of density ρ is given by τff =
[3π/(32Gρ)]1/2 = 3.4(100/nH2 )1/2 Myr, where nH2 is
the number density of H2 molecules, and the mass per H2
molecule is 3.9 × 10−24 g for a fully molecular gas of cosmological composition. This is the timescale on which an
object that experiences no significant forces other than its
own gravity will collapse to a singularity. For an object
with αG ≈ 1, the crossing timescale is τcr = R/σ ≈ 2τff .
It is of great interest how these natural timescales compare
to cloud lifetimes and depletion times.
Scoville et al. (1979) argue that GMCs in the Milky Way
are very long-lived (> 108 yr) based on the detection
of molecular clouds in interarm regions, and Koda et al.
(2009) apply similar arguments to the H2 -rich galaxy M51.
They find that, while the largest GMC complexes reside
within the arms, smaller (< 104 M⊙ ) clouds are found
in the interarm regions, and the molecular fraction is large
(> 75%) throughout the central 8 kpc (see also Foyle et al.
2010). This suggests that massive GMCs are rapidly builtup in the arms from smaller, pre-existing clouds that survive the transit between spiral arms. The massive GMCs
fragment into the smaller clouds upon exiting the arms,
but have column densities high enough to remain molecular (see § 3.4). Since the time between spiral arm passages is ∼ 100 Myr, this implies a similar cloud lifetime
τlife & 100 Myr ≫ τff . Note, however, this is an argument
for the mean lifetime of a H2 molecule, not necessarily for
a single cloud. Furthermore, these arguments do not apply
to H2 -poor galaxies like the LMC and M33.
Kawamura et al. (2009, see also Fukui et al. 1999;
Gratier et al. 2012) use the NANTEN Survey of 12 CO J=10 emission from the LMC, which is complete for clouds
with mass > 5 × 104 M⊙ , to identify three distinct cloud
types that are linked to specific phases of cloud evolution.
Type I clouds are devoid of massive star formation and represent the earliest phase. Type II clouds contain compact
H II regions, signaling the onset of massive star formation.
Type III clouds, the final stage, harbor developed stellar
clusters and H II regions. The number counts of cloud
2.6. Star Formation Rates and Efficiencies
We can also measure star formation activity within
clouds. We define the star formation efficiency or yield, ǫ∗ ,
as the instantaneous fraction of a cloud’s mass that has been
transformed into stars, ǫ∗ = M∗ /(M∗ + Mgas ), where M∗
is the mass of newborn stars. In an isolated, non-accreting
cloud, ǫ∗ increases monotonically, but in an accreting cloud
it can decrease as well. Krumholz and McKee (2005), building on classical work by Zuckerman and Evans (1974),
argue that a more useful quantity than ǫ∗ is the star formation efficiency per free-fall time, defined as ǫff =
Ṁ∗ /(Mgas /τff ), where Ṁ∗ is the instantaneous star formation rate. This definition can also be phrased in terms of
the depletion timescale introduced above: ǫff = τff /τdep .
One virtue of this definition is that it can be applied at a
range of densities ρ, by computing τff (ρ) then taking Mgas
to be the mass at a density ≥ ρ (Krumholz and Tan 2007).
As newborn stars form in the densest regions of clouds,
ǫ∗ can only increase as one increases the density threshold
used to define Mgas . It is in principle possible for ǫff to
both increase and decrease, and its behavior as a function
7
of density encodes important information about how star
formation behaves.
Within individual clouds, the best available data on ǫ∗
and ǫff come from campaigns that use the Spitzer Space
Telescope to obtain a census of young stellar objects with
excess infrared emission, a feature that persists for 2–3 Myr
of pre-main sequence evolution. These are combined with
cloud masses and surface densities measured by millimeter dust emission or infrared extinction of background stars.
For a set of five star forming regions investigated in the
Cores to Disks Spitzer Legacy program, Evans et al. (2009)
found ǫ∗ =0.03–0.06 over entire GMCs, and ǫ∗ ∼ 0.5
considering only dense gas with n ∼ 105 cm−3 . On the
other hand, ǫff ≈ 0.03 − 0.06 regardless of whether one
considers the dense gas or the diffuse gas, due to a rough
cancellation between the density dependence of Mgas and
τff . Heiderman et al. (2010) obtain comparable values in
15 additional clouds from the Gould’s Belt Survey. Murray
(2011) find significantly higher values of ǫff = 0.14 − 0.24
for the star clusters in the Galaxy that are brightest in
WMAP free-free emission, but this value may be biased
high because it is based on the assumption that the molecular clouds from which those clusters formed have undergone negligible mass loss despite the clusters’ extreme luminosities (Feldmann and Gnedin 2011).
At the scale of the Milky Way as a whole, recent estimates based on a variety of indicators put the galactic star
formation rate at ≈ 2 M⊙ yr−1 (Robitaille and Whitney
2010; Murray and Rahman 2010; Chomiuk and Povich
2011), within a factor of ∼ 2 of earlier estimates based
on ground-based radio catalogs (e.g., McKee and Williams
1997). In comparison, the total molecular mass of the Milky
Way is roughly 109 M⊙ (Solomon et al. 1987), and this,
combined with the typical free-fall time estimated in the
previous section, gives a galaxy-average ǫff ∼ 0.01 (see
also Krumholz and Tan 2007; Murray and Rahman 2010).
For extragalactic sources one can measure ǫff by combining SFR indicators such as Hα, ultraviolet, and infrared emission with tracers of gas at a variety of densities. As discussed above, observed H2 depletion times are
τdep (H2 ) ≈ 2 Gyr, whereas GMC densities of nH ∼30–
1000 cm−3 correspond to free-fall times of ∼ 1–8 Myr,
with most of the mass probably closer to the smaller value,
since the mass spectrum of GMCs ensures that most mass
is in large clouds, which tend to have lower densities.
Thus ǫff ∼0.001–0.003. Observations using tracers of
dense gas (n ∼ 105 cm−3 ) such as HCN yield ǫff ∼
0.01 (Krumholz and Tan 2007; Garcı́a-Burillo et al. 2012);
given the errors, the difference between the HCN and CO
values is not significant. As with the Evans et al. (2009)
clouds, higher density regions subtend smaller volumes and
comprise smaller masses. ǫff is nearly constant because
Mgas and 1/τff both fall with density at about the same rate.
Figure 3 shows a large sample of observations compiled by Krumholz et al. (2012), which includes individual
Galactic clouds, nearby galaxies, and high-redshift galaxies, covering a very large range of mean densities. They
Fig. 3.— SFR per unit area versus gas column density over freefall time Krumholz et al. (2012). Different shapes indicate different data sources, and colors represent different types of objects:
red circles and squares are Milky Way clouds, black filled triangles and unresolved z = 0 galaxies, black open triangles are unresolved z = 0 starbursts, blue filled symbols are unresolved z > 1
disk galaxies, and blue open symbols are unresolved z > 1 starburst galaxies. Contours show the distribution of kpc-sized regions
within nearby galaxies. The black line is ǫff = 0.01, and the gray
band is a factor of 3 range around it.
find that all of the data are consistent with ǫff ∼ 0.01, albeit
with considerable scatter and systematic uncertainty. Even
with the uncertainties, however, it is clear that ǫff ∼ 1 is
strongly ruled out.
2.7. GMCs in Varying Galactic Environments
One gains useful insight into GMC physics by studying
their properties as a function of environment. Some of the
most extreme environments, such as those in starbursts or
metal-poor galaxies, also offer unique insights into astrophysics in the primitive universe, and aid in the interpretation of observations of distant sources.
Galactic centers, which feature high metallicity and stellar density, and often high surface densities of gas and star
formation, are one unusual environment to which we have
observational access. The properties of the bulge, and presence of a bar appear to influence the amount of H2 in the
center (Fisher et al. 2013). Central regions with high ΣH2
preferentially show reduced τdep (H2 ) compared to galaxy
averages (Leroy et al. 2013), suggesting that central GMCs
convert their gas into stars more rapidly. Reduced τdep (H2 )
is correlated with an increase in CO (2-1)/(1-0) ratios, indicating enhanced excitation (or lower optical depth). Many
galaxy centers also exhibit a super-exponential increase in
CO brightness, and a drop in CO-to-H2 conversion factor
(Sandstrom et al. 2012, which reinforces the short τdep (H2 )
8
conclusion). On the other hand, in our own Galactic Center,
Longmore et al. (2013) show that there are massive molecular clouds that have surprisingly little star formation, and
depletion times τdep (H2 ) ∼ 1 Gyr comparable to disk
GMCs (Kruijssen et al. 2013), despite volume and column
densities orders of magnitude higher (see Longmore et al.
Chapter).
Obtaining similar spatially-resolved data on external
galaxies is challenging. Rosolowsky and Blitz (2005) examined several very large GMCs (M ∼ 107 M⊙ , R ∼
40 − 180 pc) in M 64. They also find a size-linewidth coefficient somewhat larger than in the Milky Way disk, and,
in 13 CO, high surface densities. Recent multi-wavelength,
high-resolution ALMA observations of the center of the
nearby starburst NGC 253 find cloud masses M ∼ 107 M⊙
and sizes R ∼ 30 pc, implying ΣGMC & 103 M⊙ pc−2
(Leroy et al. 2013, in prep.). The cloud linewidths imply
that they are self-gravitating.
The low metallicity environments of dwarf galaxies and
outer galaxy disks supply another fruitful laboratory for
study of the influence of environmental conditions. Because
of their proximity, the Magellanic Clouds provide the best
locations to study metal-poor GMCs. Owing to the scarcity
of dust at low metallicity (e.g., Draine et al. 2007) the abundances of H2 and CO in the ISM are greatly reduced compared to what would be found under comparable conditions
in a higher metallicity galaxy (see the discussion in § 3.3).
As a result, CO emission is faint, only being present in regions of very high column density (e.g., Israel et al. 1993;
Bolatto et al. 2013, and references therein). Despite these
difficulties, there are a number of studies of low metallicity GMCs. Rubio et al. (1993) reported GMCs in the SMC
exhibit sizes, masses, and a size-linewidth relation similar
to that in the disk of the Milky Way. However, more recent work suggests that GMCs in the Magellanic Clouds
are smaller and have lower masses, brightness temperatures,
and surface densities than typical inner Milky Way GMCs,
although they are otherwise similar to Milky Way clouds
(Fukui et al. 2008; Bolatto et al. 2008; Hughes et al. 2010;
Muller et al. 2010; Herrera et al. 2013). Magellanic Cloud
GMCs also appear to be surrounded by extended envelopes
of CO-faint H2 that are ∼ 30% larger than the CO-emitting
region (Leroy et al. 2007, 2009). Despite their CO faintness, though, the SFR-H2 relation appears to be independent of metallicity once the change in the CO-to-H2 conversion factor is removed (Bolatto et al. 2011).
which may be continuous or intermittent. Each mechanism
however acts over different sizes and timescales, producing density enhancements of different magnitudes. Consequently, different mechanisms may dominate in different
environments, and may lead to different cloud properties. In
addition to these physical mechanisms for gathering mass,
forming a GMC involves a phase change in the ISM, and
this too may happen in a way that depends on the largescale environment (§ 3.6)
Two processes that we will not consider as cloud formation mechanisms are thermal instabilities (TI, Field 1965)
and magneto-rotational instabilities (MRI, Balbus and Hawley
1991). Neither of these by themselves are likely to form
molecular clouds. TI produces the cold (100 K) atomic
component of the ISM, but the cloudlets formed are ∼
pc in scale, and without the shielding provided by a large
gas column this does not lead to 10 − 20 K molecular
gas. Furthermore, TI does not act in isolation, but interacts with all other processes taking place in the atomic ISM
(Vázquez-Semadeni et al. 2000; Sánchez-Salcedo et al. 2002;
Piontek and Ostriker 2004; Kim et al. 2008). Nevertheless the formation of the cold atomic phase aids GMC
formation by enhancing both shock compression and vertical settling, as discussed further in § 3.2. The MRI is
not fundamentally compressive, so again it will not by
itself lead to GMC formation, although it can significantly affect the development of large-scale gravitational
instabilities that are limited by galactic angular momentum (Kim et al. 2003). MRI or TI may also drive turbulence (Koyama and Inutsuka 2002; Kritsuk and Norman
2002; Kim et al. 2003; Piontek and Ostriker 2005, 2007;
Inoue and Inutsuka 2012), and thereby aid cloud agglomeration and contribute to converging flows. However, except
in regions with very low SFR, the amplitudes of turbulence
driven by TI and MRI are lower than those driven by star
formation feedback.
3.1 Localized Converging Flows
Stellar feedback processes such as the expansion of H II
regions (Bania and Lyon 1980; Vazquez-Semadeni et al.
1995; Passot et al. 1995) and supernova blast waves (McCray and Kafatos
1987; Gazol-Patiño and Passot 1999; de Avillez 2000;
de Avillez and Mac Low 2001; de Avillez and Breitschwerdt
2005; Kim et al. 2011; Ntormousi et al. 2011) can drive
converging streams of gas that accumulate to become
molecular clouds, either in the Galactic plane or above it,
or even after material ejected vertically by the local excess
of pressure due to the stars/supernovae falls back into the
plane of the disk. Morphological evidence for this process
can be found in large-scale extinction maps of the Galaxy
(see Fig. 4), and recent observations in both the Milky Way
and the LMC, confirm that MCs can be found at the edges
of supershells (Dawson et al. 2011, 2013).
Locally – on scales up to ∼ 100 pc – it is likely that these
processes play a dominant role in MC formation, since on
these scales the pressure due to local energy sources is typ-
3. GMC FORMATION
We now turn to the question of how GMCs form. The
main mechanisms that have been proposed, which we discuss in detail below, are converging flows driven by stellar feedback or turbulence (§ 3.1), agglomeration of smaller
clouds (§ 3.2), gravitational instability (§ 3.3) and magnetogravitational instability (§ 3.4), and instability involving differential buoyancy (§ 3.5). All these mechanisms involve
converging flows, i.e., ∇ · u < 0, where u is the velocity,
9
ically P/kB ∼ 104 K cm−3 , which exceeds the mean pressure of the ISM in the Solar neighborhood (Draine 2011).
The mass of MCs created by this process will be defined by
the mean density ρ0 and the velocity correlation length, L of
the converging streams; L is less than the disk scale height
H for local turbulence. For Solar neighborhood conditions,
where there are low densities and relatively short timescales
for coherent flows, this implies a maximum MC mass of a
few times 104 M⊙ . Converging flows driven by large-scale
instabilities can produce higher-mass clouds (see below).
divided by the inflow velocity (∼ few km s−1 ). In addition
to the overall age spread, the shape of the stellar age distribution produced by this mechanism is consistent with those
observed in nearby MCs: most of the stars have ages of 13 Myr, and only few older. While some of these regions
may be contamination by non-members (Hartmann 2003),
the presence of some older stars might be the result of a few
stars forming prior to the global but hierarchical and chaotic
contraction concentrates most of the gas and forms most of
the stars (Hartmann et al. 2012). This model is also consistent with the observed morphology of Solar neighborhood
clouds, since their elongated structures would naturally result from inflow compression.
3.2 Spiral-Arm Induced Collisions
While localized converging flows can create clouds with
masses up to ∼ 104 M⊙ , most of the molecular gas in
galaxies is found in much larger clouds (§ 2.3). Some mechanism is required to either form these large clouds directly,
or to induce smaller clouds to agglomerate into larger ones.
Although not yet conclusive, there is some evidence of different cloud properties in arms compared to inter-arm regions, which could suggest different mechanisms operatFig. 4.— Extinction map towards the Orion-Monoceros molecu- ing (Colombo et al., submitted). It was long thought that
lar complex. Different features are located at different distances,
the agglomeration of smaller clouds could not work bebut at the mean distance of Orion, the approximate size of the A
cause, for clouds moving with observed velocity disperand B clouds, as well as the λ Ori ring, is of the order of 50 pc.
sions, the timescale required to build a 105 − 106 M⊙ cloud
Note the presence of shells at a variety of scales, presumably due
would be > 100 Myr (Blitz and Shu 1980). However in
to OB stars and/or supernovae. The λ Ori ring surrounds a 5 Myr
the presence of spiral arms, collisions between clouds beold stellar cluster, and it is thought to have been produced by a
come much more frequent, greatly reducing the timescale
supernova (data from Rowles and Froebrich 2009, as adapted by
(Casoli and Combes 1982; Kwan and Valdes 1983; Dobbs
L. Hartmann (2013, in preparation)).
2008). Fig. 5 shows a result from a simulation where
A converging flow it not by itself sufficient to form GMCs predominantly form from spiral arm induced collia MC; the detailed initial velocity, density, and mag- sions. Even in the absence of spiral arms, in high surface
netic field structure must combine with TI to produce fast density galaxies Tasker and Tan (2009) find that collisions
cooling (e.g., McCray et al. 1975; Bania and Lyon 1980; are frequent enough (every ∼ 0.2 orbits) to influence GMC
Vazquez-Semadeni et al. 1995; Hennebelle and Pérault 1999; properties, and possibly star formation (Tan 2000). The freKoyama and Inutsuka 2000; Audit and Hennebelle 2005; quency and success of collisions is enhanced by clouds’
Heitsch et al. 2006; Vázquez-Semadeni et al. 2007). This mutual gravity (Kwan and Valdes 1987; Dobbs 2008), but
allows rapid accumulation of cold, dense atomic gas, and is suppressed by magnetic fields (Dobbs and Price 2008).
This mechanism can explain a number of observed
thus promotes molecule formation. The accumulation of
features
of GMCs. Giant molecular associations in spigas preferentially along field lines (perhaps due to magnetoral
arms
display a quasi-periodic spacing along the arms
Jeans instability – see § 3.4) also increases the mass to flux
(Elmegreen
and Elmegreen 1983; Efremov 1995), and
ratio, causing a transition from subcritical gas to supercritiDobbs
(2008)
show that this spacing can be set by the
cal (Hartmann et al. 2001; Vázquez-Semadeni et al. 2011).
epicyclic
frequency
imposed by the spiral perturbation,
Thus the accumulation of mass from large-scale streams,
which
governs
the
amount
of material which can be colthe development of a molecular phase with negligible therlected
during
a
spiral
arm
passage.
A stronger spiral potenmal support, and the transition from magnetically subcrititial
produces
more
massive
and
widely
spaced clouds. The
cal to supercritical all happen essentially simultaneously, at
stochasticity
of
cloud-cloud
collisions
naturally produces
21
−2
a column density of (∼ 10 cm ; Hartmann et al. 2001,
a
powerlaw
GMC
mass
function
(Field
and Saslaw 1965;
see also Section 3.6), allowing simultaneous molecular
Penston
et
al.
1969;
Taff
and
Savedoff
1973;
Handbury et al.
cloud and star formation.
1977;
Kwan
1979;
Hausman
1982;
Tomisaka
1984), and
This mechanism for GMC formation naturally explains
the
powerlaw
indices
produced
in
modern
hydrodynamic
the small age spread observed in MCs near the Sun (see
§ 2.5), since the expected star formation timescale in these simulations agree well with observations (Dobbs 2008;
models is the thickness of the compressed gas (∼ few pc) Dobbs et al. 2011a; Tasker and Tan 2009), provided the
simulations also include a subgrid feedback recipe strong
10
enough to prevent runaway cloud growth (Dobbs et al.
2011a; Hopkins et al. 2011). A third signature of either local converging flows or agglomeration is that they can produce clouds that are counter-rotating compared to the galactic disk (Dobbs 2008; Dobbs et al. 2011a; Tasker and Tan
2009), consistent with observations showing that retrograde clouds are as common as prograde ones (Blitz 1993;
Phillips 1999; Rosolowsky et al. 2003; Imara and Blitz
2011; Imara et al. 2011). Clouds formed by agglomeration
also need not be gravitationally bound, although, again,
stellar feedback is necessary to maintain an unbound population (Dobbs et al. 2011b).
Kim and Ostriker (2007); Robertson and Kravtsov (2008);
Tasker and Tan (2009); Dobbs et al. (2011a); Wada et al.
(2011); Hopkins (2012). In cases where a grand design
is present (e.g., when driven by a tidal encounter), gas
flowing through the spiral pattern supersonically experiences a shock, which raises the density and can trigger
gravitational collapse (Roberts 1969); numerical investigations of this include e.g., Kim and Ostriker (2002, 2006);
Shetty and Ostriker (2006); Dobbs et al. (2012); see also
§ 3.4 for a discussion of magnetic effects.
Clouds formed by gravitational instabilities in singlephase gas disks typically have masses ∼ 10 times the
two-dimensional Jeans mass MJ,2D = c4eff /(G2 Σ) ∼
108 M⊙ (ceff /7 km s−1 )4 (Σ/M⊙ pc−2 )−1 . The gathering
scale for mass is larger than the 2D Jeans length LJ,2D =
c2eff /(GΣ) in part because the fastest-growing scale exceeds LJ,2D even for infinitesimally-thin disks, and this increases for thick disks; also, the cloud formation process
is highly anisotropic because of shear. At moderate gas
surface densities, Σ < 100M⊙ pc−2 , as found away from
spiral-arm regions, the corresponding masses are larger than
those of observed GMCs when ceff ∼ 7 km s−1 , comparable to large-scale mean velocity dispersions for the atomic
medium. The absence of such massive clouds is consistent with observations indicating that Q values are generally
above the critical threshold (except possibly in high redshift systems), such that spiral-arm regions at high Σ and/or
processes that reduce ceff locally are required to form observed GMCs via self-gravitating instability. While many
analyses and simulations assume a single-phase medium,
the diffuse ISM from which GMCs form is in fact a multiphase medium, with cold clouds surrounded by a warmer
intercloud medium. The primary contribution to the effective velocity dispersion of the cold medium is turbulence.
This turbulence can dissipate due to cloud-cloud collisions,
as well as large-scale flows in the horizontal direction, and
flows towards the midplane from high latitude. Turbulent
dissipation reduces the effective pressure support, allowing instability at lower Σ (Elmegreen 2011). In addition,
the local reduction in ceff enables gravitational instability to
form lower mass clouds. Simulations that include a multiphase medium and/or feedback from star formation (which
drives turbulence and also breaks up massive GMCs) find
a broad spectrum of cloud masses, extending up to several ×106 M⊙ (e.g., Wada et al. 2000; Shetty and Ostriker
2008; Tasker and Tan 2009; Dobbs et al. 2011a; Hopkins
2012).
3.3 Gravitational Instability
An alternative explanation for massive clouds is that they
form in a direct, top-down manner, and one possible mechanism for this to happen is gravitational instability. Axisymmetric perturbations in single-phase infinitesimally thin gas
disks with effective sound speed ceff , surface density Σ, and
epicyclic frequency κ can occur whenever the Toomre parameter Q ≡ κceff /(πGΣ) < 1. For the nonaxisymmetric
case, however, there are no true linear (local) instabilities
because differential rotation ultimately shears any wavelet
into a tightly wrapped trailing spiral with wavenumber
k ∝ t in which pressure stabilization (contributing to the
dispersion relation as k 2 c2eff ) is stronger than self-gravity
(contributing as −2πGΣgas |k|). Using linear perturbation
theory, Goldreich and Lynden-Bell (1965b) were the first to
analyze the growth of low-amplitude shearing wavelets in
gaseous disks; the analogous “swing amplification” process
for stellar disks was studied by Julian and Toomre (1966).
Spiral-arm regions, because they have higher gaseous surface densities, are more susceptible than interarm regions
to self-gravitating fragmentation, and this process has
been analyzed using linear theory (e.g., Elmegreen 1979;
Balbus and Cowie 1985; Balbus 1988). Magnetic effects
are particularly important in spiral-arm regions because of
the reduced shear compared to interarm regions; this combination leads to a distinct process termed the magneto-Jeans
instability (MJI; see § 3.4).
Nonaxisymmetric disturbances have higher Q thresholds
for growth (i.e., amplification is possible at lower Σ) than
axisymmetric disturbances. Because the instabilities are
nonlinear, numerical simulations are required to evaluate
these thresholds. Kim and Ostriker (2001) found, for local thin-disk simulations and a range of magnetization, that
the threshold is at Q = 1.2 − 1.4. Vertical thickness of
the disk dilutes self-gravity, which tends to reduce the critical Q below unity, but magnetic fields and the contribution of stellar gravity provide compensating effects, yielding Qcrit ∼ 1.5 (Kim et al. 2002, 2003; Kim and Ostriker
2007; Li et al. 2005).
In the absence of a pre-existing “grand design” stellar spiral pattern, gravitational instability in the combined
gas-star disk can lead to flocculent or multi-armed spirals, as modeled numerically by e.g., Li et al. (2005);
3.4 Magneto-Jeans Instability
In magnetized gas disks, another process, now termed
the magneto-Jeans instability (MJI), can occur. This was
first investigated using linear perturbation theory for disks
with solid-body rotation by Lynden-Bell (1966), and subsequently by Elmegreen (1987), Gammie (1996), and
Kim and Ostriker (2001) for general rotation curves. MJI
occurs in low-shear disks at any nonzero magnetization,
11
whereas swing amplification (and its magnetized variants)
described in § 3.3 requires large shear. In MJI, magnetic tension counteracts the Coriolis forces that can otherwise suppress gravitational instability in rotating systems, and low shear is required so that self-gravity takes
over before shear increases k to the point of stabilization.
Kim and Ostriker (2001) and Kim et al. (2002) have simulated this process in uniform, low-shear regions as might be
found in inner galaxies.
Because low shear is needed for MJI, it is likely to be
most important either in central regions of galaxies or in spiral arms (Elmegreen 1994), where compression by a factor
Σ/Σ0 reduces the local shear as d ln Ω/d ln R ∼ Σ/Σ0 − 2.
Given the complex dynamics of spiral arm regions, this
is best studied with MHD simulations (Kim and Ostriker
2002, 2006; Shetty and Ostriker 2006), which show that
MJI can produce massive, self-gravitating clouds either
within spiral arms or downstream, depending on the
strength of the spiral shock. Clouds that collapse downstream are found within overdense spurs. The spiral arms
maintain their integrity better in magnetized models that in
comparison unmagnetized simulations. Spacings of spiralarm spurs formed by MJI are several times the Jeans length
c2eff /(GΣarm ) at the mean arm gas surface density Σarm ,
consistent with observations of spurs and giant H II regions that form “beads on a string” in grand design spirals
(Elmegreen and Elmegreen 1983; La Vigne et al. 2006).
The masses of gas clumps formed by MJI are typically
∼ 107 M⊙ in simulations with ceff = 7 km s−1 , comparable to the most massive observed giant molecular cloud
complexes and consistent with expectations from linear theory. Simulations of MJI in multiphase disks with feedbackdriven turbulence have not yet been conducted, so there are
no predictions for the full cloud spectrum.
simulations including self-consistent flow through the spiral arm indicate that vertical gradients in horizontal velocity
(a consequence of vertically-curved spiral shocks) limit the
development of the undular Parker mode (Kim and Ostriker
2006). While Parker instability has important consequences
for vertical redistribution of magnetic flux, and there is
strong evidence for the formation of magnetic loops anchored by GMCs in regions of high magnetic field strength
such as the Galactic Center (Fukui et al. 2006; Torii et al.
2010), it is less clear if Parker instability can create massive,
highly overdense clouds.
For a medium subject to TI, cooling of overdense
gas in magnetic valleys can strongly enhance the density contrast of structures that grow by Parker instability (Kosiński and Hanasz 2007; Mouschovias et al. 2009).
However, simulations to date have not considered the more
realistic case of a pre-existing cloud / intercloud medium
in which turbulence is also present. For a non-turbulent
medium, cold clouds could easily slide into magnetic valleys, but large turbulent velocities may prevent this. In
addition, whatever mechanisms drive turbulence may disrupt the coherent development of large-scale Parker modes.
An important task for future modeling of the multiphase,
turbulent ISM is to determine whether Parker instability
primarily re-distributes the magnetic field and alters the
distribution of low-density coronal gas, or if it also plays
a major role in creating massive, bound GMCs near the
midplane.
3.6 Conversion of H to H2 , and C+ to CO
Thus far our discussion of GMC formation has focused on the mechanisms for accumulating high density
gas. However, the actual observable that defines a GMC is
usually CO emission, or in some limited cases other tracers of H2 (e.g., Bolatto et al. 2011). Thus we must also
consider the chemical transition from atomic gas, where
the hydrogen is mostly H I and carbon is mostly C+ , to
molecular gas characterized by H2 and CO. The region
over which this transition occurs is called a photodissociation or photon-dominated region (PDR). H2 molecules
form (primarily) on the surfaces of dust grains, and are
destroyed by resonant absorption of Lyman- and Wernerband photons. The equilibrium H2 abundance is controlled
by the ratio of the far ultraviolet (FUV) radiation field
to the gas density, so that H2 becomes dominant in regions where the gas is dense and the FUV is attenuated
(van Dishoeck and Black 1986; Black and van Dishoeck
1987; Sternberg 1988; Krumholz et al. 2008, 2009a; Wolfire et al.
2010). Once H2 is abundant, it can serve as the seed
for fast ion-neutral reactions that ultimately culminate
in the formation of CO. Nonetheless CO is also subject
to photodissociation, and it requires even more shielding than H2 before it becomes the dominant C repository
(van Dishoeck and Black 1988).
The need for high extinction to form H2 and CO has
two important consequences. One, already alluded to in
3.5 Parker Instability
A final possible top-down formation mechanism is
Parker instability, in which horizontal magnetic fields
threading the ISM buckle due to differential buoyancy
in the gravitational field, producing dense gas accumulations at the midplane. Because the most unstable
modes have wavelengths ∼ (1 − 2)2πH, where H is
the disk thickness, gas could in principle be gathered
from scales of several hundred pc to produce quite massive clouds (Mouschovias 1974; Mouschovias et al. 1974;
Blitz and Shu 1980). However, Parker instabilities are selflimiting (e.g., Matsumoto et al. 1988), and in single-phase
media saturate at factor of few density enhancements at
the midplane (e.g., Basu et al. 1997; Santillán et al. 2000;
Machida et al. 2009). Simulations have also demonstrated
that three-dimensional dynamics, which enhance reconnection and vertical magnetic flux redistribution, result in end
states with relatively uniform horizontal density distributions (e.g., Kim et al. 1998, 2003). Some models suggest
that spiral arm regions may be favorable for undular modes
to dominate over interchange ones (Franco et al. 2002), but
12
§ 2, is that the transition between H I and H2 is shifted
to higher surface densities in galaxies with low metallicity
and dust abundance and thus low extinction per unit mass
(Fumagalli et al. 2010; Bolatto et al. 2011; Wong et al.
2013). A second is that the conversion factor XCO between CO emission and H2 mass increases significantly
in low metallicity galaxies (Bolatto et al. 2013, and references therein), because the shift in column density is much
greater for CO than for H2 . CO and H2 behave differently
because self-shielding against dissociation is much stronger
for H2 than for CO, which must rely on dust shielding
(Wolfire et al. 2010). As a result, low metallicity galaxies
show significant CO emission only from high-extinction
peaks of the H2 distribution, rather than from the bulk of
the molecular material.
While the chemistry of GMC formation is unquestionably important for understanding and interpreting observations, the connection between chemistry and dynamics is considerably less clear. In part this is due to
numerical limitations. At the resolutions achievable in
galactic-scale simulations, one can model H2 formation
only via subgrid models that are either purely analytic
(e.g., Krumholz et al. 2008, 2009a,b; McKee and Krumholz
2010; Narayanan et al. 2011, 2012; Kuhlen et al. 2012;
Jaacks et al. 2013) or that solve the chemical rate equations with added “clumping factors” that are tuned to reproduce observations (e.g., Robertson and Kravtsov 2008;
Gnedin et al. 2009; Pelupessy and Papadopoulos 2009;
Christensen et al. 2012). Fully-time-dependent chemodynamical simulations that do not rely on such methods
are restricted to either simple low-dimensional geometries
(Koyama and Inutsuka 2000; Bergin et al. 2004) or to regions smaller than typical GMCs (Glover and Mac Low
2007a,b; Glover et al. 2010; Glover and Mac Low 2011;
Clark et al. 2012; Inoue and Inutsuka 2012).
One active area of research is whether the H I to H2
or C+ to CO transitions are either necessary or sufficient
for star formation. For CO the answer appears to be
“neither”. The nearly fixed ratio of CO surface brightness to star formation rate we measure in Solar metallicity regions (see § 2.1) drops dramatically in low metallicity
ones (Gardan et al. 2007; Wyder et al. 2009; Bolatto et al.
2011, 2013), strongly suggesting that the loss of metals
is changing the carbon chemistry but not the way stars
form. Numerical simulations suggest that, at Solar metallicity, CO formation is so rapid that even a cloud undergoing free-fall collapse will be CO-emitting by the time
there is substantial star formation (Hartmann et al. 2001;
Heitsch and Hartmann 2008a). Conversely, CO can form
in shocks even if the gas is not self-gravitating (Dobbs et al.
2008; Inoue and Inutsuka 2012). Moreover, formation of
CO does not strongly affect the temperature of molecular clouds, so it does not contribute to the loss of thermal support and onset of collapse (Krumholz et al. 2011;
Glover and Clark 2012a). Thus it appears that CO accompanies star formation, but is not causally related to it.
The situation for H2 is much less clear. Unlike CO,
the correlation between star formation-H2 correlation appears to be metallicity-independent, and is always stronger
than the star formation-total gas correlation. Reducing
the metallicity of a galaxy at fixed gas surface density
lower both the H2 abundance and star formation rate,
and by nearly the same factor (Wolfe and Chen 2006;
Rafelski et al. 2011; Bolatto et al. 2011). This still does
not prove causality, however. Krumholz et al. (2011) and
Glover and Clark (2012a) suggest that the explanation is
that both H2 formation and star formation are triggered by
shielding effects. Only at large extinction is the photodissociation rate suppressed enough to allow H2 to become
dominant, but the same photons that dissociate H2 also heat
the gas via the grain photoelectric effect, and gas only gets
cold enough to form stars where the photoelectric effect is
suppressed. This is, however, only one possible explanation
of the data.
A final question is whether the chemical conditions in
GMCs are in equilibrium or non-equilibrium. The gasphase ion-neutral reactions that lead to CO formation have
high rates even at low temperatures, so carbon chemistry is likely to be in equilibrium (Glover et al. 2010;
Glover and Mac Low 2011; Glover and Clark 2012b). The
situation for H2 is less clear. The rate coefficient for H2
formation on dust grain surfaces is quite low, so whether
gas can reach equilibrium depends on the density structure and the time available. Averaged over ∼ 100 pc
size scales and at metallicities above ∼ 1% of Solar,
Krumholz and Gnedin (2011) find that equilibrium models of Krumholz et al. (2009a) agree very well with timedependent ones by Gnedin et al. (2009), and the observed
metallicity-dependence of the H I-H2 transition in external galaxies is also consistent with the predictions of the
equilibrium models (Fumagalli et al. 2010; Bolatto et al.
2011; Wong et al. 2013). However, Mac Low and Glover
(2012) find that equilibrium models do not reproduce their
simulations on ∼ 1 − 10 pc scales. Nonetheless, observations of Solar Neighborhood clouds on such scales appear to be consistent with equilbrium (Lee et al. 2012). All
models agree, however, that non-equilibrium effects must
become dominant at metallicities below ∼ 1 − 10% of
Solar, due to the reduction in the rate coefficient for H2 formation that accompanies the loss of dust (Krumholz 2012;
Glover and Clark 2012b).
4. STRUCTURE, EVOLUTION, AND DESTRUCTION
Now that we have sketched out how GMCs come into
existence, we consider the processes that drive their internal
structure, evolution, and eventual dispersal.
4.1 GMC Internal Structure
GMCs are characterized by a very clumpy, filamentary structure (see the reviews by André et al. and
13
Molinari et al. in this volume), which can be produced by a wide range of processes, including gravitational collapse (e.g., Larson 1985; Nagai et al. 1998; Curry
2000; Burkert and Hartmann 2004), non-self-gravitating
supersonic turbulence (Padoan et al. 2001), and colliding flows plus thermal instability (Audit and Hennebelle
2005; Vázquez-Semadeni et al. 2006); all these processes
can be magnetized or unmagnetized. Some recent observational studies argue that the morphology is consistent
with multi-scale infall (e.g., Galván-Madrid et al. 2009;
Schneider et al. 2010; Kirk et al. 2013), but strong conclusions will require quantitative comparison with a wide
range of simulations.
One promising approach to such comparisons is to
develop statistical measures that can be applied to both
simulations and observations, either two-dimensional column density maps or three-dimensional position-positionvelocity cubes. Padoan et al. (2004b,a) provide one example. They compare column density PDFs produced
in simulations of sub-Alfvénic and super-Alfvénic turbulence, and argue that observed PDFs are better fit by the
super-Alfvénic model. This is in some tension with observations showing that magnetic fields remain well-ordered
over a wide range of length scales (see the recent review
by Crutcher 2012 and Li et al., this volume). The need
for super-Alfvénic turbulence in the simulations may arise
from the fact that they did not include self-gravity, thus
requiring stronger turbulence to match the observed level
of structure (Vázquez-Semadeni et al. 2008). Nevertheless, the Padoan et al. results probably do show that magnetic fields cannot be strong enough to render GMCs subAlfvénic.
A second example comes from Brunt (2010), Brunt et al.
(2010a,b) and Price et al. (2011), who use the statistics
of the observed 2D column density PDF to infer the underlying 3D volume density PDF, and in turn use this
to constrain the relationship between density variance
and Mach number in nearby molecular clouds. They
conclude from this analysis that a significant fraction
of the energy injection that produces turbulence must
be in compressive rather than solenoidal modes. Various authors (Kainulainen et al. 2009a; Kritsuk et al. 2011;
Ballesteros-Paredes et al. 2011a; Federrath and Klessen
2013) also argue that the statistics of the density field are
also highly sensitive to the amount of star formation that
has taken place in a cloud, and can therefore be used as a
measure of evolutionary state.
high value of the Reynolds number essentially guarantees
that the flow will be turbulent. Moreover, since the bulk
velocity greatly exceeds the sound speed, the turbulence
must be supersonic, though not necessarily super-Alfvénic.
Zuckerman and Evans (1974) proposed that this turbulence
would be confined to small scales, but modern simulations
of supersonic turbulence indicate that the power is mostly
on large scales. It is also possible that the linewidths contain a significant contribution from coherent infall, as we
discuss below.
While turbulence is expected, the deeper question is why
the linewidths are so large in the first place. Simulations
conducted over the last ∼ 15 years have generally demonstrated that, in the absence of external energy input, turbulence decays in ∼ 1 crossing time of the outer scale of
the turbulent flow (Mac Low et al. 1998; Mac Low 1999;
Stone et al. 1998; Padoan and Nordlund 1999), except in
the case of imbalanced MHD turbulence (Cho et al. 2002).
Thus the large linewidths observed in GMCs would not in
general be maintained for long periods in the absence of
some external input. This problem has given rise to a number of proposed solutions, which can be broadly divided
into three categories: global collapse, externally-driven turbulence, and internally-driven turbulence.
4.2.1 The Global Collapse Scenario
The global collapse scenario, first proposed by Goldreich and Kwan
(1974) and Liszt et al. (1974), and more recently revived by
Vázquez-Semadeni et al. (2007, 2009), Heitsch and Hartmann
(2008a,b), Heitsch et al. (2009), Ballesteros-Paredes et al.
(2011b,a), and Hartmann et al. (2012) as a nonlinear version of the hierarchical fragmentation scenario proposed
by Hoyle (1953), offers perhaps the simplest solution: the
linewidths are dominated by global gravitational collapse
rather than random turbulence. This both provides a natural
energy source (gravity) and removes the need to explain
why the linewidths do not decay, because in this scenario
GMCs, filaments, and clumps are not objects that need to
be supported, but rather constitute a hierarchy of stages in
a global, highly inhomogeneous collapse flow, with each
stage accreting from its parent (Vázquez-Semadeni et al.
2009).
Investigations of this scenario generally begin by considering an idealized head-on collision between two singlephase, warm, diffuse gas streams, which might be caused
by either local feedback or large-scale gravitational instability (cf. § 3). (Simulations of more realistic glancing collisions between streams already containing dense clumps
have yet to be performed.) The large scale compression
triggers the formation of a cold cloud, which quickly acquires many Jeans masses because the warm-cold phase
transition causes a factor of ∼ 100 increase in density
and a decrease by the same factor in temparature. Thus
the Jeans mass, MJ ∝ ρ−1/2 T 3/2 , decreases by a factor ∼ 104 (e.g., Vazquez-Semadeni 2012). The cloud
therefore readily fragments into clumps (Heitsch et al.
4.2 Origin of Nonthermal Motions
As discussed in § 2.3, GMCs contain strong nonthermal motions, with the bulk of the energy in modes with
size scales comparable to the cloud as a whole. For a typical GMC density of ∼ 100 cm−3 , temperature of ∼ 10
K, and bulk velocity of ∼ 1 km s−1 , the viscous dissipation scale is ∼ 1012 cm (∼0.1 AU), implying that the
Reynolds number of these motions is ∼ 109 . Such a
14
2005; Vázquez-Semadeni et al. 2006), and the ensemble collapse. This would naturally explain relatively long GMC
of clumps becomes gravitationally unstable and begins lifetimes and (as we discuss in § 5) low star formation rates,
an essentially pressure-free collapse. It contracts first but in turn raises the problem of why these motions do not
along its shortest dimension (Lin et al. 1965), producing decay, giving rise to a global collapse. The external driving
sheets and then filaments (Burkert and Hartmann 2004; scenario proposes that this decay is offset by the injection of
Hartmann and Burkert 2007; Vázquez-Semadeni et al. 2007, energy by flows external to the GMC. One obvious source
2010, 2011; Heitsch and Hartmann 2008a; Heitsch et al. of such external energy is the accretion flow from which the
2009). Although initially the motions in the clouds are cloud itself forms, which can be subject to non-linear thin
random and turbulent, they become ever-more infall- shell instability (Vishniac 1994) or oscillatory overstability
dominated as the collapse proceeds. However, because (Chevalier and Imamura 1982) that will drive turbulence.
these motions have a gravitational origin, they natu- Klessen and Hennebelle (2010) point out that only a small
rally appear virialized (Vázquez-Semadeni et al. 2007; fraction of the gravitational potential energy of material acBallesteros-Paredes et al. 2011b). Accretion flows consis- creting onto a GMC would need to be converted to bulk motent with the scenario have been reported in several observa- tion before it is dissipated in shocks in order to explain the
tional studies of dense molecular gas (e.g., Galván-Madrid et al.observed linewidths of GMCs, and semi-analytic models by
2009; Schneider et al. 2010; Kirk et al. 2013), but observa- Goldbaum et al. (2011) confirm this conclusion. Numerical
tions have yet to detect the predicted inflows at the early, simulations confirm that cold dense layers confined by the
large-scale stages of the hierarchy. These are difficult to ram pressure of accretion flows indeed are often turbulent
detect because it is not easy to separate the atomic medium (Hunter et al. 1986; Stevens et al. 1992; Walder and Folini
directly connected to molecular clouds from the general 2000; Koyama and Inutsuka 2002; Audit and Hennebelle
H I in the galaxy, and because the GMCs are highly frag- 2005; Heitsch et al. 2005; Vázquez-Semadeni et al. 2006),
mented, blurring the inverse p-Cygni profiles expected for although numerical simulations consistently show that
infall. In fact, Heitsch et al. (2009) show that the CO line the velocity dispersions of these flows are significantly
profiles of chaotically collapsing clouds match observations smaller than those observed in GMCs unless the flows are
of GMCs.
self-gravitating (Koyama and Inutsuka 2002; Heitsch et al.
While this idea elegantly resolves the problem of the 2005; Vázquez-Semadeni et al. 2007, 2010). This can be
large linewidths, it faces challenges with respect to the understood because the condition of simultaneous therconstraints imposed by the observed 20 − 30 Myr life- mal and ram pressure balance implies that the Mach numtimes of GMCs (§ 2.5) and the low rates and efficiencies bers in both the warm and cold phases are comparable
of star formation (§ 2.6). We defer the question of star (Banerjee et al. 2009).
formation rates and efficiencies to § 5. Concerning GMC
While accretion flows are one possible source of energy,
lifetimes, a semi-analytic model by Zamora-Avilés et al. there are also others. Galactic-scale and kpc-scale simula(2012) for the evolution of the cloud mass and SFR in tions by Tasker and Tan (2009), Tasker (2011), Dobbs et al.
this scenario shows agreement with the observations within (2011a,b, 2012), Dobbs and Pringle (2013), Hopkins et al.
factors of a few. Slightly smaller timescales (∼ 10 Myr) (2012), Van Loo et al. (2013) all show that GMCs are emare observed in numerical simulations of cloud build-up bedded in large-scale galactic flows that subject them to
that consider the evolution of the molecular content of the continuous external buffeting – cloud-cloud collisions, encloud (Heitsch and Hartmann 2008a), although these au- counters with shear near spiral arms, etc. – even when
thors considered substantially smaller cloud masses and the cloud’s mass is not necessarily growing. These extermore dense flows. Simulations considering larger cloud nal motions are particularly important for the most massive
masses (several 104 M⊙ ) exhibit evolutionary timescales clouds, which preferentialy form via large-scale galactic
∼ 20 Myr (Colı́n et al. 2013), albeit no tracking of the flows, and can drive turbulence in them over a time signifmolecular fraction was performed there. Simulations based icantly longer than τff . This mechanism seems particularly
on the global collapse scenario have not yet examined the likely to operate in high-surface density galaxies where the
formation and evolution of clouds of masses above 105 M⊙ , entire ISM is molecular and thus there is no real distinction
comparable to those studied in extragalactic observations. between GMCs and other gas, and in fact seems to be reMoreover, in all of these models the lifetime depends crit- quired to explain the large velocity dispersions observed in
ically on the duration and properties of the gas inflow that high-redshift galaxies (Krumholz and Burkert 2010).
forms the clouds, which is imposed as a boundary condition. Self-consistent simulations in which the required in- 4.2.3 The Internal Driving Scenario
flows are generated by galactic scale flows also remain to
be performed.
The internally-driven scenario proposes that stellar feedback internal to a molecular cloud is responsible for driving
4.2.2 The External Driving Scenario
turbulence and explaining the large linewidths seen, in conjunction with externally-driven turbulence in the very rare
The alternative possibility is that the large linewidths of clouds without significant star formation (e.g., the so-called
GMCs are dominated by random motions rather than global Maddalena’s cloud; Williams et al. 1994). There are a num15
y (kpc)
log H2 dz [cm]
ber of possible sources of turbulent driving, including H II
5
regions, radiation pressure, protostellar outflows, and the
19.4
winds of main sequence stars. (Supernovae are unlikely to
be important as an internal driver of turbulence in GMCs
19.2
in most galaxies because the stellar evolution timescale is
comparable to the crossing timescale, though they could be
important as an external driver, for the dispersal of GMCs,
19
0
and in starburst galaxies, see below). Matzner (2002) and
Dekel and Krumholz (2013) provide useful summaries of
the momentum budgets associated with each of these mech18.8
anisms, and they are discussed in much more detail in the
chapter by Krumholz et al. in this volume.
18.6
H II regions are one possible turbulent driver. Krumholz et al.
(2006) and Goldbaum et al. (2011) conclude from their
-5
-5
0
5
semi-analytic models that H II regions provide a power
x (kpc)
source sufficient to offset the decay of turbulence, that most
of this power is distributed into size scales comparable to Fig. 5.— A snapshot is shown from a simulation of the gas in
the size of the cloud, and that feedback power is comparable a spiral galaxy, including a fixed sprial potential, gas heating and
to accretion power. The results from simulations are less cooling, self gravity and stellar feedback (from Dobbs and Pringle
clear. Gritschneder et al. (2009) and Walch et al. (2012) 2013). The GMCs in this simulation predominanlty form by
find that H II regions in their simulations drive turbulence at cloud-cloud collisions in the spiral arms. The black and white
velocity dispersions comparable to observed values, while color scheme shows the total gas column density, whilst the blueDale et al. (2005, 2012, 2013) and Colı́n et al. (2013) find yellow scheme shows the H2 fraction integrated through the disc.
that H II regions rapidly disrupt GMCs with masses up to
∼ 105 M⊙ within less than 10 Myr (consistent with observations showing that > 10 Myr-old star clusters are usually
gas-free – Leisawitz et al. 1989; Mayya et al. 2012), but do
not drive turbulence. The origin of the difference is not
clear, as the simulations differ in several ways, including
the geometry they assume, the size scales they consider,
and the way that they set up the initial conditions.
Nevertheless, in GMCs where the escape speed approaches the 10 km s−1 sound speed in photoionized
gas, H II regions can no longer drive turbulence nor disrupt the clouds, and some authors have proposed that radiation pressure might take over (Thompson et al. 2005;
Krumholz and Matzner 2009; Fall et al. 2010; Murray et al.
2010; Hopkins et al. 2011, 2012). Simulations on this point
are far more limited, and the only ones published so far
that actually include radiative transfer (as opposed to a subgrid model for radiation pressure feedback) are those of
Krumholz and Thompson (2012, 2013), who conclude that
radiation pressure is unlikely to be important on the scales
of GMCs. Figure 6 shows a result from one of these simulations.
Stellar winds and outflows can also drive turbulence.
Winds from hot main-sequence stars have been studied Fig. 6.— Time slice from a radiation-hydrodyanmic simulation
by Dale and Bonnell (2008) and Rogers and Pittard (2013), of a molecular cloud with a strong radiation flux passing through
who find that they tend to ablate the clouds but not to drive it. In the top panel, color shows density and arrows show velocity;
significant turbulence. On the other hand, studies of proto- in the bottom panel, color shows temperature and arrows show
stellar outflows find that on small scales they can maintain radiation flux. Density and temperature are both normalized to
a constant velocity dispersion while simultaneously keep- characteristic values at the cloud edge; velocity is normalized to
ing the star formation rate low (Cunningham et al. 2006; the sound speed at the cloud edge, and flux to the injected radiation
flux. The simulation demonstrates that radiation pressure-driven
Li and Nakamura 2006; Matzner 2007; Nakamura and Li
turbulence is possible, but also that the required radiation flux and
2007; Wang et al. 2010). On the other hand, these studies matter column density are vastly in excess of the values found in
also indicate that outflows cannot be the dominant driver of real GMCs. Figure taken from Krumholz and Thompson (2012).
turbulence on ∼ 10 − 100 pc scales in GMCs, both because
16
they lack sufficient power, and because they tend to produce
a turbulent power spectrum with a distinct bump at ∼ 1 pc
scales, in contrast to the pure powerlaw usually observed.
Whether any of these mechanisms can be the dominant
source of the large linewidths seen in GMCs remains unsettled. One important caveat is that only a few of the
simulations with feedback have included magnetic fields,
and Wang et al. (2010) and Gendelev and Krumholz (2012)
show (for protostellar outflows and H II regions, respectively) that magnetic fields can dramatically increase the
ability of internal mechanisms to drive turbulence, because
they provide an effective means of transmitting momentum
into otherwise difficult-to-reach portions of clouds. MHD
simulations of feedback in GMCs are clearly needed.
4.3 Mass Loss and Disruption
As discussed in § 2.5, GMCs are disrupted long before they can turn a significant fraction of their mass into
stars. The question of what mechanism or mechanisms
are responsible for this is closely tied to the question of
the origin of GMC turbulence discussed in the previous
section, as each proposed answer to that question imposes
certain requirements on how GMCs must disrupt. In the
global collapse scenario, disruption must occur in less than
the mean-density free-fall time to avoid excessive star formation. Recent results suggest a somewhat slower collapse in flattened or filamentary objects (Toalá et al. 2012;
Pon et al. 2012), but disruption must still be fast. In the
externally-driven or internally-driven turbulence scenarios
disruption can be slower, but must still occur before the
bulk of the material can be converted to stars. Radiation
pressure and protostellar outflows appear unlikely to be
responsible, for the same reasons (discussed in the previous section) that they cannot drive GMC-scale turbulence.
For main sequence winds, Dale and Bonnell (2008) and
Rogers and Pittard (2013) find that they can expel mass
from small, dense molecular clumps, but it is unclear if the
same is true of the much larger masses and lower density
scales that characterize GMCs.
The remaining stellar feedback mechanisms, photoionization and supernovae, are more promising. Analytic models have long predicted that photoionization should be the
primary mechanism for ablating mass from GMCs (e.g.,
Field 1970; Whitworth 1979; Cox 1983; Williams and McKee
1997; Matzner 2002; Krumholz et al. 2006; Goldbaum et al.
2011), and numerical simulations by Dale et al. (2012,
2013) and Colı́n et al. (2013) confirm that photoionization
is able to disrupt GMCs with masses of ∼ 105 M⊙ or less.
More massive clouds, however, may have escape speeds too
high for photoionization to disrupt them unless they suffer
significant ablation first.
Supernovae are potentially effective in clouds of all
masses, but are in need of further study. Of cataloged
Galactic supernova remnants, 8% (and 25% of X-ray emitting remnants) are classified as “mixed morphology,” believed to indicate interaction between the remnant and
17
dense molecular gas (Rho and Petre 1998). This suggests that a non-negligible fraction of GMCs may interact with supernovae. Because GMCs are clumpy, this interaction will differ from the standard solutions in a uniform medium (e.g., Cioffi et al. 1988; Blondin et al. 1998),
but theoretical studies of supernova remnants in molecular gas have thus far focused mainly on emission properties (e.g., Chevalier 1999; Chevalier and Fransson 2001;
Tilley et al. 2006) rather than the dynamical consequences
for GMC evolution. Although some preliminary work
(Kovalenko and Korolev 2012) suggests that an outer shell
will still form, with internal clumps accelerated outward
when they are overrun by the expanding shock front (cf.
Klein et al. 1994; Mac Low et al. 1994), complete simulations of supernova remant expansion within realistic GMC
enviroments are lacking. Obtaining a quantitative assessment of the kinetic energy and momentum imparted to the
dense gas will be crucial for understanding GMC destruction.
5. Regulation of star formation in GMCs
Our discussion thus far provides the framework to address the final topic of this review: what are the dominant
interstellar processes that regulate the rate of star formation
at GMC and galactic scales? The accumulation of GMCs
is the first step in star formation, and large scale, top-down
processes appear to determine a cloud’s starting mean density, mass to magnetic flux ratio, Mach number, and boundedness. But are these initial conditions retained for times
longer than a cloud dynamical time, and do they affect the
formation of stars within the cloud? If so, how stars form
is ultimately determined by the large scale dynamics of the
host galaxy. Alternatively, if the initial state of GMCs is
quickly erased by internally-driven turbulence or external
perturbations, then the regulatory agent of star formation
lies instead on small scales within individual GMCs. In
this section, we review the proposed schemes and key GMC
properties that regulate the production of stars.
5.1 Regulation Mechanisms
Star formation occurs at a much lower pace than its theoretical possible free-fall maximum (see Section 2.6). Explaining why this is so is a key goal of star formation theories. These theories are intimately related to the assumptions made about the evolutionary path of GMCs. Two theoretical limits for cloud evolution are a state of global collapse with a duration ∼ τff and a quasi-steady state in which
clouds are supported for times ≫ τff .
In the global collapse limit, one achieves low SFRs by
having a low net star formation efficiency ǫ∗ over the lifetime ∼ τff of any given GMC, and then disrupting the
GMC via feedback. The mechanisms invoked to accomplish this are the same as those invoked in Section 4.2.3 to
drive internal turbulence: photoionization and supernovae.
Some simulations suggest this these mechanisms can indeed enforce low ǫ∗ : Vázquez-Semadeni et al. (2010) and
Colı́n et al. (2013), using a subgrid model for ionizing feedback, find that ǫ∗ . 10% for clouds up to ∼ 105 M⊙ ,
and Zamora-Avilés et al. (2012) find that the evolutionary
timescales produced by this mechanism of cloud disruption are consistent with those inferred in the Large Magellanic Cloud (Section 2.5). On the other hand, it remains
unclear what mechanisms might be able to disrupt ∼ 106
M⊙ clouds.
If clouds are supported against large-scale collapse,
then star formation consists of a small fraction of the
mass “percolating” through this support to collapse and
form stars. Two major forms of support have been considered: magnetic (e.g., Shu et al. 1987; Mouschovias
1991a,b) and turbulent (e.g., Mac Low and Klessen 2004;
Ballesteros-Paredes et al. 2007). While dominant for over
two decades, the magnetic support theories, in which
the percolation was allowed by ambipolar diffusion, are
now less favored, (though see Mouschovias et al. 2009;
Mouschovias and Tassis 2010) due to growing observational evidence that molecular clouds are magnetically supercritical (Section 2.4). We do not discuss these models
further.
In the turbulent support scenario, supersonic turbulent
motions behave as a source of pressure with respect to
structures whose size scales are larger than the largest
scales of the turbulent motions (the “energy containing
scale” of the turbulence), while inducing local compressions at scales much smaller than that. A simple analytic
argument suggests that, regardless of whether turbulence is
internally- or externally-driven, its net effect is to increase
2
, where vrms
the effective Jeans mass as MJ,turb ∝ vrms
is the rms turbulent velocity (Mac Low and Klessen 2004).
Early numerical simulations of driven turbulence in isothermal clouds (Klessen et al. 2000; Vázquez-Semadeni et al.
2003) indeed show that, holding all other quantities fixed,
raising the Mach number of the flow decreases the dimensionless star formation rate ǫff . However, this is true
only as long as the turbulence is maintained; if it is allowed to decay, then raising the Mach number actually
raises ǫff , because in this case the turbulence simply accelerates the formation of dense regions and then dissipates (Nakamura and Li 2005). Magnetic fields, even
those not strong enough to render the gas subcritical, also
decrease ǫff (Heitsch et al. 2001; Vázquez-Semadeni et al.
2005; Padoan and Nordlund 2011; Federrath and Klessen
2012).
To calculate the SFR in this scenario, one can idealize
the turbulence level, mean cloud density, and SFR as quasistationary, and then attempt to compute ǫff . In recent years,
a number of analytic models have been developed to do so
(Krumholz and McKee 2005; Padoan and Nordlund 2011;
Hennebelle and Chabrier 2011; see Federrath and Klessen
2012 for a useful compilation, and for generalizations of
several of the models). These models generally exploit the
fact that supersonic isothermal turbulence produces a probability density distribution (PDF) with a lognormal form
(Vazquez-Semadeni 1994), so that there is always a fraction
of the mass at high densities. The models then assume that
the mass at high densities (above some threshold), Mhd , is
responsible for the instantaneous SFR, which is given as
SFR=Mhd /τ , where τ is some characteristic timescale of
the collapse at those high densities.
In all of these models ǫff is determined by other dimensionelss numbers: the rms turbulent Mach number M,
the virial ratio αG , and (when magnetic fields are considered) the magnetic β parameter; the ratio of compressive
to solenoidal modes in the turbulence is a fourth possible
parameter (Federrath et al. 2008; Federrath and Klessen
2012). The models differ in their choices of density threshold and timescale (see the chapter by Padoan et al.), leading
to variations in the predicted dependence of ǫff on M, αG ,
and β. However, all the models produce ǫff ∼ 0.01 − 0.1
for dimensionless values comparable to those observed.
Federrath and Klessen (2012) and Padoan et al. (2012)
have conducted large campaigns of numerical simulations where they have systematically varied M, αG , and
β, measured ǫff , and compared to the analytic models.
Padoan et al. (2012) give their results in terms of the ratio tff /tdyn rather than αG , but the two are identical up to
a constant factor (Tan et al. 2006). In general they find that
ǫff decreases strongly with αG and increases weakly with
M, and that a dynamically-significant magnetic field (but
not one so strong as to render the gas subcritical) reduces ǫff
by a factor of ∼ 3. Simulations produce ǫff ∼ 0.01 − 0.1,
in general agreement with the range of analytic predictions.
One can also generalize the quasi-stationary turbulent
support models by embedding them in time-dependent
models for the evolution of a cloud as a whole. In this approach one computes the instantaneous SFR from a cloud’s
current state (using one of the turbulent support models
or based on some other calibration from simulations), but
the total mass, mean density, and other quantities evolve in
time, so that the instantaneous SFR does too. In this type
of model, a variety of assumptions are necessarily made
about the cloud’s geometry and about the effect of the stellar feedback. Krumholz et al. (2006) and Goldbaum et al.
(2011) adopt a spherical geometry and compute the evolution from the virial theorem, assuming that feedback
can drive turbulence that inhibits collapse. As illustrated
in Figure 7, they find that most clouds undergo oscillations around equilibrium before being destroyed at final
SFEs ∼ 5–10%. The models match a wide range of observations, including the distributions of column density,
linewidth-size relation, and cloud lifetime. In constrast,
Zamora-Avilés et al. (2012, also shown in Figure 7) adopt a
planar geometry (which implies longer free-fall times than
in the spherical case; Toalá et al. 2012) and assume that
feedback does not drive turbulence or inhibit contraction.
With these models they reproduce the star formation rates
seen in low- and high-mass clouds and clumps, and the
stellar age distributions in nearby clusters. As shown in the
Figure, the overall evolution is quite different in the two
models, with the Goldbaum et al. clouds undergoing multiple oscillations at roughly fixed Σ and M∗ /Mgas , while the
18
imply that there is a cut-off of ΣSFR at low gas surface densities, although simple models of gravitational instability in
isothermal disks can indeed reproduce this result (Li et al.
2005), but instead may indicate that additional parameters
beyond just Σgas control ΣSFR . In outer disks, the ISM
is mostly diffuse atomic gas and the radial scale length of
Σgas is quite large (comparable to the size of the optical
disk; Bigiel and Blitz (2012)). The slow fall-off of Σgas
with radial distance implies that the sensitivity of ΣSFR to
other parameters will become more evident in these regions.
For example, a higher surface density in the old stellar disk
appears to raise ΣSFR (Blitz and Rosolowsky 2004, 2006;
Leroy et al. 2008), likely because stellar gravity confines
the gas disk, raising the density and lowering the dynamical
time. Conversely, ΣSFR is lower in lower-metallicity galaxies (Bolatto et al. 2011), likely because lower dust shielding
against UV radiation inhibits the formation of a cold, starforming phase (Krumholz et al. 2009b).
Feedback must certainly be part of this story. Recent
large scale simulations of disk galaxies have consistently
pointed to the need for feedback to prevent runaway collapse and limit star formation rates to observed levels (e.g.,
Kim et al. 2011; Tasker 2011; Hopkins et al. 2011, 2012;
Dobbs et al. 2011a; Shetty and Ostriker 2012; Agertz et al.
2013). With feedback parameterizations that yield realistic SFRs, other ISM properties (including turbulence levels and gas fractions in different H I phases) are also realistic (see above and also Joung et al. 2009; Hill et al.
2012). However, it still also an open question whether
feedback is the entire story for the large scale SFR. In some
simulations (e.g., Ostriker and Shetty 2011; Dobbs et al.
2011a; Hopkins et al. 2011, 2012; Shetty and Ostriker
2012; Agertz et al. 2013), the SFR on & 100 pc scales
is mainly set by the time required for gas to become
gravitationally-unstable on large scales and by the parameters that control stellar feedback, and is insensitive to the
parameterization of star formation on . pc scales. In
other models the SFR is sensitive to the parameters describing both feedback and small-scale star formation (e.g.,
ǫff and H2 chemistry; Gnedin and Kravtsov 2010, 2011;
Kuhlen et al. 2012, 2013).
Part of this disagreement is doubtless due to the fact
that current simulations do not have sufficient resolution
to include the details of feedback, and in many cases they
do not even include the required physical mechanisms (for
example radiative transfer and ionization chemistry). Instead, they rely on subgrid models for momentum and energy injection by supernovae, radiation, and winds, and
the results depend on the details of how these mechanisms
are implemented. Resolving the question of whether feedback alone is sufficient to explain the large-scale star formation rate of galaxies will require both refinement of
the subgrid feedback models using high resolution simulations, and comparison to observations in a range of environments. In at least some cases, the small-scale simulations have raised significant doubts about popular subgrid
models (e.g., Krumholz and Thompson 2012, 2013).
3.0
2.5
2.0
1.5
1.0
0.5
Goldbaum+
0.0
Zamora-Avilés+
−0.5
1
0
−1
−2
Type I
−3
Type II
−4
Type III
−5
0 10 20 30 40 50 60 70 80
t [Myr]
log
∗
M /M
gas
log Σgas [
M
⊙
pc−2 ]
Zamora-Avilés et al. model predicts a much more monotonic evolution. Differentiating between these two pictures
will require a better understanding of the extent to which
feedback is able to inhibit collapse.
Fig. 7.— Predictions for the large-scale evolution of GMCs using the models of Goldbaum et al. (2011, thin lines, each line corresponding to a different realization of a stochastic model) and
Zamora-Avilés et al. (2012, thick line). The top panel shows the
gas surface density. The minimum in the Goldbaum et al. models is the threshold at which CO dissociates. For the planar
Zamora-Avilés et al. model, the thick line is the median and the
shaded region is the 10th - 90th percentile range for random orientation. The bottom panel shows the ratio of instantaneous stellar to
gas mass. Colors indicate the type following the Kawamura et al.
(2009) classification (see Section 2.5), computed based on the Hα
and V -band luminosities of the stellar populations.
5.2 Connection Between Local and Global Scales
Extraglactic star formation observations at large scales
average over regions several times the disk scale height
in width, and over many GMCs. As discussed in Section
2.1, there is an approximately linear correlation between
the surface densities of SFR and molecular gas in regions
where Σgas . 100M⊙ pc−2 , likely because observations
are simply counting the number of GMCs in a beam. At
higher Σgas , the volume filling factor of molecular material approaches unity, and the index N of the correlation
ΣSFR ∝ ΣN
gas increases. This can be due to increasing density of molecular gas leading to shorter gravitational collapse and star formation timescales, or because higher total
gas surface density leads to stronger gravitational instability and thus faster star formation. At the low values of Σgas
found in the outer disks of spirals (and in dwarfs), the index N is also greater than unity. This does not necessarily
19
A number of authors have also developed analytic models for large-scale star formation rates in galactic disks.
Krumholz et al. (2009b) propose a model in which the fraction of the ISM in a star-forming molecular phase is determined by the balance between photodissociation and H2
formation, and the star formation rate within GMCs is determined by the turbulence-regulated star formation model
of Krumholz and McKee (2005). This model depends on
assumed relations between cloud complexes and the properties of the interstellar medium on large scales, including
the assumption that the surface density of cloud complexes
is proportional to that of the ISM on kpc scales, and that the
mass fraction in the warm atomic ISM is negligible compared to the mass in cold atomic and molecular phases.
Ostriker et al. (2010) and Ostriker and Shetty (2011) have
developed models in which star formation is self-regulated
by feedback. In these models, the equilibrium state is
found by simultaneously balancing ISM heating and cooling, turbulent driving and dissipation, and gravitational confinement with pressure support in the diffuse ISM. The
SFR adjusts to a value required to maintain this equilibrium state. Numerical simulations by Kim et al. (2011) and
Shetty and Ostriker (2012) show that ISM models including
turbulent and radiative heating feedback from star formation indeed reach the expected self-regulated equilibrium
states. However, as with other large-scale models, these
simulations rely on subgrid feedback recipes whose accuracy have yet to be determined. In all of these models,
in regions where most of the neutral ISM is in gravitationally bound GMCs, ΣSF R depends on the internal state
of the clouds through the ensemble average of ǫff /τff . If
GMC internal states are relatively independent of their environments, this would yield values of hǫff /τff i that do not
strongly vary within a galaxy or from one galaxy to another,
naturally explaining why τdep (H2 ) appears to be relatively
uniform, ∼ 2 Gyr wherever Σgas . 100M⊙ pc−2 .
Many of the recent advances in understanding largescale star formation have been based on disk galaxy systems similar to our own Milky Way. Looking to the future,
we can hope that the methods being developed to connect
individual star-forming GMCs with the larger scale ISM in
local “laboratories” will inform and enable efforts in highredshift systems, where conditions are more extreme and
observational constraints are more challenging.
resolving GMC complexes and investigating the physical
state and formation mechanisms of GMCs in a variety of
extragalactic environments. The smaller interferometers
like CARMA and SMA will likely focus on systematic
mapping of large areas in the Milky Way or even external
galaxies. At cm wavelengths, the recently upgraded JVLA
brings new powerful capabilities in continuum detection at
7 mm to study cool dust in disks, as well as the study of
free-free continuum, molecular emission, and radio recombination lines in our own galaxy and other galaxies. Singledish mm-wave facilities equipped with array receivers and
continuum cameras such as the IRAM 30m, NRO 45m,
LMT 50m, and the future CCAT facility will enable fast
mapping of large areas in the Milky Way, providing the
much needed large scale context to high resolution studies. These facilities, along with the APEX and NANTEN2
telescopes, will pursue multitransition / multiscale surveys
sampling star forming cores and the parent cloud material
simultaneously and providing valuable diagnostics of gas
kinematics and physical conditions.
There is hope that some of the new observational capabilities will break the theoretical logjam described in this
review. At present, it is not possible for observations to distinguish between the very different cloud formation mechanisms proposed in Section 3, nor between the mechanisms
that might be responsible for controlling cloud density and
velocity structure, and cloud disruption (Section 4), nor between various models for how star formation is regulated
(Section 5). There is reason to hope that the new data that
will become available in the next few years will start to rule
some of these models out.
6.2 Simulations and theory
The developments in numerical simulations of GMCs since
PPV will continue over the next few years to PPVII. There
is a current convergence of simulations towards the scales
of GMCs, and GMC scale physics. Galaxy simulations
are moving towards ever smaller scales, whilst individual
cloud simulations seek to include more realistic initial conditions. At present, galactic models are limited by the resolution required to adequately capture stellar feedback, internal cloud structure, cloud motions and shocks. They also do
not currently realise the temperatures or densities of GMCs
observed, nor typically include magnetic fields. Smaller
scale simulations, on the other hand, miss larger scale dynamics such as spiral shocks, shear and cloud–cloud collisions, which will be present either during the formation or
throughout the evolution of a molecular cloud. Future simulations, which capture the main physics on GMC scales,
will provide a clearer picture of the evolution and lifetimes
of GMCs. Furthermore they will be more suitable for determining the role of different processes in driving turbulence and regulating star formation in galaxies, in conjunction with analytic models and observations.
Continuing improvements in codes, including the use of
moving mesh codes in addition to AMR and SPH methods,
6. Looking forward
6.1 Observations
Systematic surveys and detailed case studies will be enriched by the expansion in millimeter-submillimeter capabilities over the next decade. These will contribute in
two main modes: through cloud-scale observations in other
galaxies, and in expanding the study of clouds in our own
Milky Way. The increased sensitivities of ALMA and
NOEMA will sample smaller scales at larger distances,
20
will also promote progress on GMCs descriptions, and allow more consistency checks between different numerical
methods. More widespread, and further development of
chemodynamical modelling will enable the study of different tracers in cloud and galaxy simulations. In conjunction
with these techniques, synthetic observations, such as HI
and CO maps, will become increasingly important for comparing the results of numerical models and simulations, and
testing whether the simulations are indeed viable representations of galaxies and clouds.
Bertoldi F. and McKee C. F. (1992) Astrophys. J., 395, 140.
Bigiel F. and Blitz L. (2012) Astrophys. J., 756, 183.
Bigiel F. et al. (2008) Astron. J., 136, 2846.
Black J. H. and van Dishoeck E. F. (1987) Astrophys. J., 322, 412.
Blair G. N. et al. (1978) Astrophys. J., 219, 896.
Blanc G. A. et al. (2009) Astrophys. J., 704, 842.
Blitz L. (1993) in: Protostars and Planets III, (edited by E. H.
Levy and J. I. Lunine), pp. 125–161.
Blitz L. and Rosolowsky E. (2004) Astrophys. J. Lett., 612, L29.
Blitz L. and Rosolowsky E. (2006) Astrophys. J., 650, 933.
Blitz L. and Shu F. H. (1980) Astrophys. J., 238, 148.
Blitz L. and Thaddeus P. (1980) Astrophys. J., 241, 676.
Blondin J. M. et al. (1998) Astrophys. J., 500, 342.
Bolatto A. D. et al. (2008) Astrophys. J., 686, 948.
Bolatto A. D. et al. (2011) Astrophys. J., 741, 12.
Bolatto A. D. et al. (2013) Annu. Rev. Astron. Astrophys., 51, 207.
Brand J. and Wouterloot J. G. A. (1995) Astron. Astrophys., 303,
851.
Brunt C. M. (2003) Astrophys. J., 584, 293.
Brunt C. M. (2010) Astron. Astrophys., 513, A67.
Brunt C. M. et al. (2009) Astron. Astrophys., 504, 883.
Brunt C. M. et al. (2010a) Mon. Not. R. Astron. Soc., 405, L56.
Brunt C. M. et al. (2010b) Mon. Not. R. Astron. Soc., 403, 1507.
Burkert A. and Hartmann L. (2004) Astrophys. J., 616, 288.
Burkert A. and Hartmann L. (2013) Astrophys. J., 773, 48.
Burton W. B. and Gordon M. A. (1978) Astron. Astrophys., 63, 7.
Calzetti D. et al. (2012) Astrophys. J., 752, 98.
Carruthers G. (1970) Astrophys. J. Lett., 161, L81.
Casoli F. and Combes F. (1982) Astron. Astrophys., 110, 287.
Cheung A. C. et al. (1968) Phys. Rev. Lett., 25, 1701.
Cheung A. C. et al. (1969) Nature, 221, 626.
Chevalier R. A. (1999) Astrophys. J., 511, 798.
Chevalier R. A. and Fransson C. (2001) Astrophys. J. Lett., 558,
L27.
Chevalier R. A. and Imamura J. N. (1982) Astrophys. J., 261, 543.
Cho J. et al. (2002) Astrophys. J., 564, 291.
Chomiuk L. and Povich M. S. (2011) Astron. J., 142, 197.
Christensen C. et al. (2012) Mon. Not. R. Astron. Soc., 425, 3058.
Cioffi D. F. et al. (1988) Astrophys. J., 334, 252.
Clark P. C. et al. (2012) Mon. Not. R. Astron. Soc., 424, 2599.
Colı́n P. et al. (2013) Mon. Not. R. Astron. Soc., 435, 1701.
Colombo D. et al. (2013) Astrophys. J., submitted.
Cox D. P. (1983) Astrophys. J. Lett., 265, L61.
Crutcher R. M. (2012) Annu. Rev. Astron. Astrophys., 50, 29.
Cunningham A. J. et al. (2006) Astrophys. J., 653, 416.
Curry C. L. (2000) Astrophys. J., 541, 831.
Dale J. E. and Bonnell I. A. (2008) Mon. Not. R. Astron. Soc., 391,
2.
Dale J. E. et al. (2005) Mon. Not. R. Astron. Soc., 358, 291.
Dale J. E. et al. (2012) Mon. Not. R. Astron. Soc., 424, 377.
Dale J. E. et al. (2013) Mon. Not. R. Astron. Soc., 430, 234.
Dame T. M. et al. (1987) Astrophys. J., 322, 1, 706.
Dame T. M. et al. (2001) Astrophys. J., 547, 1, 792.
Dawson J. R. et al. (2011) Astrophys. J., 728, 127.
Dawson J. R. et al. (2013) Astrophys. J., 763, 56.
de Avillez M. A. (2000) Mon. Not. R. Astron. Soc., 315, 479.
de Avillez M. A. and Breitschwerdt D. (2005) Astron. Astrophys.,
436, 585.
de Avillez M. A. and Mac Low M.-M. (2001) Astrophys. J. Lett.,
551, L57.
Dekel A. and Krumholz M. R. (2013) Mon. Not. R. Astron. Soc.,
432, 455.
Acknowledgments. Support for this work was provided by: the European Research Council through the
FP7 ERC starting grant project LOCALSTAR (CLD);
the NSF through grants AST09-55300 (MRK), AST1109395 (M-MML), AST09-08185 (ECO), AST09-55836
(ADB), and AST10-09049 (MH); NASA through ATP
grant NNX13AB84G (MRK), Chandra award number
GO2-13162A (MRK) issued by the Chandra X-ray Observatory Center, which is operated by the Smithsonian Astrophysical Observatory for and on behalf of the National
Aeronautics Space Administration under contract NAS803060, and Hubble Award Number 13256 (MRK) issued by
the Space Telescope Science Institute, which is operated by
the Association of Universities for Research in Astronomy,
Inc., under NASA contract NAS 5-26555; a Research Corporation for Science Advancement Cottrell Scholar Award
(ADB); an Alfred P. Sloan Fellowship (MRK); the hospitality of the Aspen Center for Physics, which is supported
by the National Science Foundation Grant PHY-1066293
(MRK); and CONACYT grant 102488 (EVS).
REFERENCES
Agertz O. et al. (2013) Astrophys. J., 770, 25.
Audit E. and Hennebelle P. (2005) Astron. Astrophys., 433, 1.
Balbus S. A. (1988) Astrophys. J., 324, 60.
Balbus S. A. and Cowie L. L. (1985) Astrophys. J., 297, 61.
Balbus S. A. and Hawley J. F. (1991) Astrophys. J., 376, 214.
Ballesteros-Paredes J. (2006) Mon. Not. R. Astron. Soc., 372, 443.
Ballesteros-Paredes J. and Hartmann L. (2007) Rev. Mexicana Astron. Astrofis., 43, 123.
Ballesteros-Paredes J. and Mac Low M.-M. (2002) Astrophys. J.,
570, 734.
Ballesteros-Paredes J. et al. (1999) Astrophys. J., 515, 286.
Ballesteros-Paredes J. et al. (2007) Protostars and Planets V, pp.
63–80.
Ballesteros-Paredes J. et al. (2011a) Mon. Not. R. Astron. Soc.,
416, 1436.
Ballesteros-Paredes J. et al. (2011b) Mon. Not. R. Astron. Soc.,
411, 65.
Bally J. et al. (1987) Astrophys. J. Lett., 312, L45.
Banerjee R. et al. (2009) Mon. Not. R. Astron. Soc., 398, 1082.
Bania T. M. and Lyon J. G. (1980) Astrophys. J., 239, 173.
Bash F. N. et al. (1977) Astrophys. J., 217, 464.
Basu S. et al. (1997) Astrophys. J. Lett., 480, L55.
Bayet E. et al. (2009) Mon. Not. R. Astron. Soc., 399, 264.
Bergin E. A. et al. (2004) Astrophys. J., 612, 921.
21
Gnedin N. Y. and Kravtsov A. V. (2010) Astrophys. J., 714, 287.
Gnedin N. Y. and Kravtsov A. V. (2011) Astrophys. J., 728, 88.
Gnedin N. Y. et al. (2009) Astrophys. J., 697, 55.
Goldbaum N. J. et al. (2011) Astrophys. J., 738, 101.
Goldreich P. and Kwan J. (1974) Astrophys. J., 189, 441.
Goldreich P. and Lynden-Bell D. (1965a) Mon. Not. R. Astron.
Soc., 130, 125.
Goldreich P. and Lynden-Bell D. (1965b) Mon. Not. R. Astron.
Soc., 130, 125.
Goldsmith P. F. and Li D. (2005) Astrophys. J., 622, 938.
Gordon M. A. and Burton W. B. (1976) Astrophys. J., 208, 1, 346.
Gould R. J. and Salpeter E. E. (1963) Astrophys. J., 138, 393.
Gratier P. et al. (2012) Astron. Astrophys., 542, A108.
Gritschneder M. et al. (2009) Astrophys. J. Lett., 694, L26.
Gutermuth R. A. et al. (2011) Astrophys. J., 739, 2, 84.
Handbury M. J. et al. (1977) Astron. Astrophys., 61, 443.
Hartmann L. (2003) Astrophys. J., 585, 398.
Hartmann L. and Burkert A. (2007) Astrophys. J., 654, 988.
Hartmann L. et al. (2001) Astrophys. J., 562, 852.
Hartmann L. et al. (2012) Mon. Not. R. Astron. Soc., 420, 1457.
Hausman M. A. (1982) Astrophys. J., 261, 532.
Heiderman A. et al. (2010) Astrophys. J., 723, 2, 1019.
Heitsch F. and Hartmann L. (2008a) Astrophys. J., 689, 290.
Heitsch F. and Hartmann L. (2008b) Astrophys. J., 689, 290.
Heitsch F. et al. (2001) Astrophys. J., 547, 280.
Heitsch F. et al. (2005) Astrophys. J. Lett., 633, L113.
Heitsch F. et al. (2006) Astrophys. J., 648, 1052.
Heitsch F. et al. (2009) Astrophys. J., 704, 1735.
Hennebelle P. and Chabrier G. (2011) Astrophys. J. Lett., 743, L29.
Hennebelle P. and Pérault M. (1999) Astron. Astrophys., 351, 309.
Herrera C. N. et al. (2013) Astron. Astrophys., in press, arXiv
1304.0453.
Heyer M. et al. (2009) Astrophys. J., 699, 1092.
Heyer M. H. and Brunt C. M. (2004) Astrophys. J. Lett., 615, L45.
Heyer M. H. et al. (1987) Astrophys. J., 321, 855.
Heyer M. H. et al. (2001) Astrophys. J., 551, 2, 852.
Hill A. S. et al. (2012) Astrophys. J., 750, 104.
Hollenbach D. and Salpeter E. E. (1971) Astrophys. J., 163, 155.
Hopkins P. F. (2012) Mon. Not. R. Astron. Soc., 423, 2016.
Hopkins P. F. et al. (2011) Mon. Not. R. Astron. Soc., 417, 950.
Hopkins P. F. et al. (2012) Mon. Not. R. Astron. Soc., 421, 3488.
Hoyle F. (1953) Astrophys. J., 118, 513.
Hughes A. et al. (2010) Mon. Not. R. Astron. Soc., 406, 3, 2065.
Hunter Jr. J. H. et al. (1986) Astrophys. J., 305, 309.
Imara N. and Blitz L. (2011) Astrophys. J., 732, 78.
Imara N. et al. (2011) Astrophys. J., 732, 79.
Inoue T. and Inutsuka S.-i. (2012) Astrophys. J., 759, 35.
Israel F. P. et al. (1993) Astron. Astrophys., 276, 25.
Jaacks J. et al. (2013) Astrophys. J., 766, 94.
Jeffries R. D. et al. (2011) Mon. Not. R. Astron. Soc., 418, 1948.
Jenkins E. B. and Tripp T. M. (2011) Astrophys. J., 734, 65.
Joung M. R. et al. (2009) Astrophys. J., 704, 137.
Julian W. H. and Toomre A. (1966) Astrophys. J., 146, 810.
Juneau S. et al. (2009) Astrophys. J., 707, 1217.
Kainulainen J. et al. (2009a) Astron. Astrophys., 508, L35.
Kainulainen J. et al. (2009b) Astron. Astrophys., 497, 399.
Kawamura A. et al. (2009) Astrophys. J. Suppl., 184, 1, 1.
Kennicutt R. C. J. (1989) Astrophys. J., 344, 685.
Kennicutt R. C. J. (1998) Astrophys. J.v.498, 498, 541.
Kennicutt R. C. J. et al. (2007) Astrophys. J., 671, 1, 333.
Kim C.-G. et al. (2008) Astrophys. J., 681, 1148.
Kim C.-G. et al. (2011) Astrophys. J., 743, 25.
Dobashi K. et al. (1996) Astrophys. J., 466, 1, 282.
Dobbs C. L. (2008) Mon. Not. R. Astron. Soc., 391, 844.
Dobbs C. L. and Price D. J. (2008) Mon. Not. R. Astron. Soc., 383,
497.
Dobbs C. L. and Pringle J. E. (2013) Mon. Not. R. Astron. Soc.,
432, 653.
Dobbs C. L. et al. (2008) Mon. Not. R. Astron. Soc., 389, 1097.
Dobbs C. L. et al. (2011a) Mon. Not. R. Astron. Soc., 417, 1318.
Dobbs C. L. et al. (2011b) Mon. Not. R. Astron. Soc., 413, 2935.
Dobbs C. L. et al. (2012) Mon. Not. R. Astron. Soc., 425, 2157.
Donovan Meyer J. et al. (2013) Astrophys. J., 772, 107.
Draine B. T. (2011) Physics of the Interstellar and Intergalactic
Medium.
Draine B. T. et al. (2007) Astrophys. J., 663, 2, 866.
Efremov Y. N. (1995) Astron. J., 110, 2757.
Elmegreen B. G. (1979) Astrophys. J., 231, 372.
Elmegreen B. G. (1987) Astrophys. J., 312, 626.
Elmegreen B. G. (1994) Astrophys. J., 433, 39.
Elmegreen B. G. (2000) Astrophys. J., 530, 277.
Elmegreen B. G. (2011) Astrophys. J., 737, 10.
Elmegreen B. G. and Elmegreen D. M. (1983) Mon. Not. R. Astron. Soc., 203, 31.
Evans N. J. I. et al. (2009) Astrophys. J. Suppl., 181, 2, 321.
Fall S. M. et al. (2010) Astrophys. J. Lett., 710, L142.
Federrath C. and Klessen R. S. (2012) Astrophys. J., 761, 156.
Federrath C. and Klessen R. S. (2013) Astrophys. J., 763, 51.
Federrath C. et al. (2008) Astrophys. J. Lett., 688, L79.
Feldmann R. and Gnedin N. Y. (2011) Astrophys. J. Lett., 727,
L12+.
Field G. et al. (2011) Mon. Not. R. Astron. Soc., 416, 1, 710.
Field G. B. (1965) Astrophys. J., 142, 531.
Field G. B. (1970) Memoires of the Societe Royale des Sciences
de Liege, 19, 29.
Field G. B. and Saslaw W. C. (1965) Astrophys. J., 142, 568.
Fisher D. B. et al. (2013) Astrophys. J., 764, 2, 174.
Foyle K. et al. (2010) Astrophys. J., 725, 1, 534.
Franco J. et al. (2002) Astrophys. J., 570, 647.
Fukui Y. and Kawamura A. (2010) Annu. Rev. Astron. Astrophys.,
48, 547.
Fukui Y. et al. (1999) PASJ, 51, 745.
Fukui Y. et al. (2006) Science, 314, 106.
Fukui Y. et al. (2008) Astrophys. J.Supplement Series, 178, 1, 56.
Fumagalli M. et al. (2010) Astrophys. J., 722, 919.
Galván-Madrid R. et al. (2009) Astrophys. J., 706, 1036.
Gammie C. F. (1996) Astrophys. J., 462, 725.
Gao Y. and Solomon P. M. (2004) Astrophys. J., 606, 1, 271.
Garcı́a-Burillo S. et al. (2012) Astron. Astrophys., 539, A8.
Gardan E. et al. (2007) Astron. Astrophys., 473, 91.
Gazol-Patiño A. and Passot T. (1999) Astrophys. J., 518, 748.
Gendelev L. and Krumholz M. R. (2012) Astrophys. J., 745, 158.
Gibson D. et al. (2009) Astrophys. J., 705, 123.
Glover S. C. O. and Clark P. C. (2012a) Mon. Not. R. Astron. Soc.,
421, 9.
Glover S. C. O. and Clark P. C. (2012b) Mon. Not. R. Astron. Soc.,
426, 377.
Glover S. C. O. and Mac Low M.-M. (2007a) Astrophys. J. Suppl.,
169, 239.
Glover S. C. O. and Mac Low M.-M. (2007b) Astrophys. J., 659,
1317.
Glover S. C. O. and Mac Low M.-M. (2011) Mon. Not. R. Astron.
Soc., 412, 337.
Glover S. C. O. et al. (2010) Mon. Not. R. Astron. Soc., 404, 2.
22
Leroy A. K. et al. (2009) Astron. J., 137, 6, 4670.
Leroy A. K. et al. (2013) Astrophys. J.Letters, 769, 1, L12.
Leroy A. K. et al. (2013) Astron. J., 146, 19.
Li Y. et al. (2005) Astrophys. J., 626, 823.
Li Z.-Y. and Nakamura F. (2006) Astrophys. J. Lett., 640, L187.
Lin C. C. et al. (1965) Astrophys. J., 142, 1431.
Liszt H. S. et al. (1974) Astrophys. J., 190, 557.
Liu G. et al. (2011) Astrophys. J., 735, 1, 63.
Lizano S. and Shu F. (1989) Astrophys. J., 342, 1, 834.
Lombardi M. et al. (2010) Astron. Astrophys., 519, L7+.
Longmore S. N. et al. (2013) Mon. Not. R. Astron. Soc., 429, 2,
987.
Lynden-Bell D. (1966) The Observatory, 86, 57.
Mac Low M. (1999) Astrophys. J., 524, 169.
Mac Low M. et al. (1998) Physical Review Letters, 80, 2754.
Mac Low M.-M. and Glover S. C. O. (2012) Astrophys. J., 746,
135.
Mac Low M.-M. and Klessen R. S. (2004) Reviews of Modern
Physics, 76, 125.
Mac Low M.-M. et al. (1994) Astrophys. J., 433, 757.
Machida M. et al. (2009) PASJ, 61, 411.
Magnani L. et al. (1985) Astrophys. J., 295, 402.
Matsumoto R. et al. (1988) PASJ, 40, 171.
Matzner C. D. (2002) Astrophys. J., 566, 302.
Matzner C. D. (2007) Astrophys. J., 659, 1394.
Mayya Y. D. et al. (2012) in: European Physical Journal Web
of Conferences, vol. 19 of European Physical Journal Web of
Conferences, p. 8006.
McCray R. and Kafatos M. (1987) Astrophys. J., 317, 190.
McCray R. et al. (1975) Astrophys. J., 196, 565.
McCrea W. H. and McNally D. (1960) Mon. Not. R. Astron. Soc.,
121, 238.
McKee C. F. and Krumholz M. R. (2010) Astrophys. J., 709, 308.
McKee C. F. and Williams J. P. (1997) Astrophys. J., 476, 144.
McKellar A. (1940) PASP, 52, 1, 187.
Miura R. E. et al. (2012) Astrophys. J., 761, 37.
Mouschovias T. (1987) in: Physical processes in interstellar
clouds; Proceedings of the NATO Advanced Study Institute, pp.
453–489, Dordrecht, D. Reidel Publishing Co.
Mouschovias T. and Spitzer L. (1976) Astrophys. J., 210, 1, 326.
Mouschovias T. C. (1974) Astrophys. J., 192, 37.
Mouschovias T. C. (1991a) in: NATO ASIC Proc. 342: The
Physics of Star Formation and Early Stellar Evolution, (edited
by C. J. Lada and N. D. Kylafis), p. 61.
Mouschovias T. C. (1991b) in: NATO ASIC Proc. 342: The
Physics of Star Formation and Early Stellar Evolution, (edited
by C. J. Lada and N. D. Kylafis), p. 449.
Mouschovias T. C. and Tassis K. (2010) Mon. Not. R. Astron. Soc.,
409, 801.
Mouschovias T. C. et al. (1974) Astron. Astrophys., 33, 73.
Mouschovias T. C. et al. (2009) Mon. Not. R. Astron. Soc., 397,
14.
Muller E. et al. (2010) Astrophys. J., 712, 2, 1248.
Müller H. S. P. et al. (2005) Journal of Molecular Structure, 742,
215.
Murray N. (2011) Astrophys. J., 729, 133.
Murray N. and Rahman M. (2010) Astrophys. J., 709, 424.
Murray N. et al. (2010) Astrophys. J., 709, 191.
Myers P. C. B. P. J. (1983) Astrophys. J., 266, 1, 309.
Nagai T. et al. (1998) Astrophys. J., 506, 306.
Nakamura F. and Li Z.-Y. (2005) Astrophys. J., 631, 411.
Nakamura F. and Li Z.-Y. (2007) Astrophys. J., 662, 395.
Kim J. et al. (1998) Astrophys. J. Lett., 506, L139.
Kim J.-h. et al. (2013) Astrophys. J., 779, 8.
Kim W.-T. and Ostriker E. C. (2001) Astrophys. J., 559, 70.
Kim W.-T. and Ostriker E. C. (2002) Astrophys. J., 570, 132.
Kim W.-T. and Ostriker E. C. (2006) Astrophys. J., 646, 213.
Kim W.-T. and Ostriker E. C. (2007) Astrophys. J., 660, 1232.
Kim W.-T. et al. (2002) Astrophys. J., 581, 1080.
Kim W.-T. et al. (2003) Astrophys. J., 599, 1157.
Kirk H. et al. (2013) Astrophys. J., 766, 115.
Klein R. I. et al. (1994) Astrophys. J., 420, 213.
Klessen R. S. and Hennebelle P. (2010) Astron. Astrophys., 520,
A17+.
Klessen R. S. et al. (2000) Astrophys. J., 535, 887.
Koda J. et al. (2009) Astrophys. J. Lett., 700, L132.
Kosiński R. and Hanasz M. (2007) Mon. Not. R. Astron. Soc., 376,
861.
Kovalenko I. and Korolev V. (2012) Conference poster, URL:
http://zofia.sese.asu.edu/∼evan/
cosmicturbulence posters/Kovalenko poster.pdf.
Koyama H. and Inutsuka S.-I. (2000) Astrophys. J., 532, 980.
Koyama H. and Inutsuka S.-i. (2002) Astrophys. J. Lett., 564, L97.
Kramer C. et al. (1998) Astron. Astrophys., 329, 249.
Kritsuk A. G. and Norman M. L. (2002) Astrophys. J. Lett., 569,
L127.
Kritsuk A. G. et al. (2011) Astrophys. J. Lett., 727, L20.
Kruijssen J. M. D. et al. (2013) ArXiv e-prints, arXiv:1306.5285.
Krumholz M. R. (2012) Astrophys. J., 759, 9.
Krumholz M. R. and Burkert A. (2010) Astrophys. J., 724, 895.
Krumholz M. R. and Gnedin N. Y. (2011) Astrophys. J., 729, 36.
Krumholz M. R. and Matzner C. D. (2009) Astrophys. J., 703,
1352.
Krumholz M. R. and McKee C. F. (2005) Astrophys. J., 630, 250.
Krumholz M. R. and Tan J. C. (2007) Astrophys. J., 654, 304.
Krumholz M. R. and Thompson T. A. (2007) Astrophys. J., 669,
289.
Krumholz M. R. and Thompson T. A. (2012) Astrophys. J., 760,
155.
Krumholz M. R. and Thompson T. A. (2013) Mon. Not. R. Astron.
Soc., 434, 2329.
Krumholz M. R. et al. (2006) Astrophys. J., 653, 361.
Krumholz M. R. et al. (2008) Astrophys. J., 689, 865.
Krumholz M. R. et al. (2009a) Astrophys. J., 693, 216.
Krumholz M. R. et al. (2009b) Astrophys. J., 699, 850.
Krumholz M. R. et al. (2011) Astrophys. J., 731, 25.
Krumholz M. R. et al. (2012) Astrophys. J., 745, 69.
Kuhlen M. et al. (2012) Astrophys. J., 749, 36.
Kuhlen M. et al. (2013) Astrophys. J., 776, 34.
Kutner M. L. et al. (1977) Astrophys. J., 215, 1, 521.
Kwan J. (1979) Astrophys. J., 229, 567.
Kwan J. and Valdes F. (1983) Astrophys. J., 271, 604.
Kwan J. and Valdes F. (1987) Astrophys. J., 315, 92.
La Vigne M. A. et al. (2006) Astrophys. J., 650, 818.
Lada C. (1976) Astrophys. J.Supplement, 32, 1, 603.
Lada C. J. et al. (2010) Astrophys. J., 724, 1, 687.
Lada C. J. et al. (2012) Astrophys. J., 745, 2, 190.
Larson R. B. (1981) Mon. Not. R. Astron. Soc., 194, 809.
Larson R. B. (1985) Mon. Not. R. Astron. Soc., 214, 379.
Lee M.-Y. et al. (2012) Astrophys. J., 748, 75.
Leisawitz D. et al. (1989) Astrophys. J. Suppl., 70, 731.
Leroy A. et al. (2005) Astrophys. J., 625, 2, 763.
Leroy A. et al. (2007) Astrophys. J., 658, 2, 1027.
Leroy A. K. et al. (2008) Astron. J., 136, 2782.
23
Santillán A. et al. (2000) Astrophys. J., 545, 353.
Schneider N. et al. (2010) Astron. Astrophys., 520, A49.
Schneider N. et al. (2012) Astron. Astrophys., 540, 1, L11.
Schneider N. et al. (2013) Astrophys. J., 766, 2, L17.
Schruba A. et al. (2010) Astrophys. J., 722, 2, 1699.
Schruba A. et al. (2011) Astron. J., 142, 2, 37.
Scoville N. Z. et al. (1979) in: The Large-Scale Characteristics
of the Galaxy, vol. 84 of IAU Symposium, (edited by W. B.
Burton), pp. 277–282.
Scoville N. Z. et al. (1987) Astrophys. J.Supplement Series (ISSN
0067-0049), 63, 821.
Scoville N. Z. S. P. M. (1975) Astrophys. J. Lett., 199, 2, L105.
Shetty R. and Ostriker E. C. (2006) Astrophys. J., 647, 997.
Shetty R. and Ostriker E. C. (2008) Astrophys. J., 684, 978.
Shetty R. and Ostriker E. C. (2012) Astrophys. J., 754, 2.
Shetty R. et al. (2011) Mon. Not. R. Astron. Soc., 412, 3, 1686.
Shetty R. et al. (2013) Mon. Not. R. Astron. Soc., 430, 1, 288.
Shu F. H. et al. (1987) Annu. Rev. Astron. Astrophys., 25, 23.
Silk J. (1980) in: Saas-Fee Advanced Course 10: Star Formation,
(edited by A. Maeder and L. Martinet), p. 133.
Snell R. L. et al. (1984) Astrophys. J., 276, 625.
Snyder L. E. et al. (1969) Phys. Rev. Lett., 22, 679.
Solomon P. M. et al. (1987) Astrophys. J., 319, 730.
Spitzer Jr. L. (1949) Astrophys. J., 109, 337.
Sternberg A. (1988) Astrophys. J., 332, 400.
Stevens I. R. et al. (1992) Astrophys. J., 386, 265.
Stone J. M. et al. (1998) Astrophys. J. Lett., 508, L99.
Stutzki J. et al. (1988) Astrophys. J., 332, 379.
Swings P. and Rosenfeld L. (1937) Astrophys. J., 86, 483.
Taff L. G. and Savedoff M. P. (1973) Mon. Not. R. Astron. Soc.,
164, 357.
Tan J. C. (2000) Astrophys. J., 536, 173.
Tan J. C. et al. (2006) Astrophys. J. Lett., 641, L121.
Tasker E. J. (2011) Astrophys. J., 730, 11.
Tasker E. J. and Tan J. C. (2009) Astrophys. J., 700, 358.
Thompson T. A. et al. (2005) Astrophys. J., 630, 167.
Tilley D. A. et al. (2006) Mon. Not. R. Astron. Soc., 371, 1106.
Toalá J. A. et al. (2012) Astrophys. J., 744, 190.
Tomisaka K. (1984) PASJ, 36, 457.
Tomisaka K. (1986) PASJ, 38, 95.
Torii K. et al. (2010) PASJ, 62, 1307.
van de Hulst H. C. (1948) Harvard Observatory Monographs, 7,
73.
van Dishoeck E. F. and Black J. H. (1986) Astrophys. J.Supplement
Series (ISSN 0067-0049), 62, 109.
van Dishoeck E. F. and Black J. H. (1988) Astrophys. J., 334, 771.
van Dishoeck E. F. and Blake G. A. (1998) Annu. Rev. Astron.
Astrophys., 36, 317.
Van Loo S. et al. (2013) Astrophys. J., 764, 36.
Vazquez-Semadeni E. (1994) Astrophys. J., 423, 681.
Vazquez-Semadeni E. (2012) ArXiv e-prints, arXiv:1208.4132.
Vazquez-Semadeni E. et al. (1995) Astrophys. J., 441, 702.
Vázquez-Semadeni E. et al. (2000) Astrophys. J., 540, 271.
Vázquez-Semadeni E. et al. (2003) Astrophys. J. Lett., 585, L131.
Vázquez-Semadeni E. et al. (2005) Astrophys. J. Lett., 630, L49.
Vázquez-Semadeni E. et al. (2006) Astrophys. J., 643, 245.
Vázquez-Semadeni E. et al. (2007) Astrophys. J., 657, 870.
Vázquez-Semadeni E. et al. (2008) Mon. Not. R. Astron. Soc., 390,
769.
Vázquez-Semadeni E. et al. (2009) Astrophys. J., 707, 1023.
Vázquez-Semadeni E. et al. (2010) Astrophys. J., 715, 1302.
Vázquez-Semadeni E. et al. (2011) Mon. Not. R. Astron. Soc., 414,
Nakano T. (1978) PASJ, 30, 1, 681.
Narayanan D. et al. (2008a) Astrophys. J., 684, 996.
Narayanan D. et al. (2008b) Astrophys. J. Lett., 681, L77.
Narayanan D. et al. (2011) Mon. Not. R. Astron. Soc., 418, 664.
Narayanan D. et al. (2012) Mon. Not. R. Astron. Soc., 421, 3127.
Narayanan G. et al. (2008c) Astrophys. J. Suppl., 177, 341.
Ntormousi E. et al. (2011) Astrophys. J., 731, 13.
Oka T. et al. (2001) Astrophys. J., 562, 1, 348.
Onodera S. et al. (2010) Astrophys. J.Letters, 722, 2, L127.
Oort J. H. (1954) Bull. Astron. Inst. Netherlands, 12, 177.
Ostriker E. C. and Shetty R. (2011) Astrophys. J., 731, 41.
Ostriker E. C. et al. (2010) Astrophys. J., 721, 975.
Padoan P. and Nordlund Å. (1999) Astrophys. J., 526, 279.
Padoan P. and Nordlund Å. (2011) Astrophys. J., 730, 40.
Padoan P. et al. (2001) Astrophys. J., 553, 227.
Padoan P. et al. (2004a) Physical Review Letters, 92, 19, 191102.
Padoan P. et al. (2004b) Astrophys. J. Lett., 604, L49.
Padoan P. et al. (2012) Astrophys. J. Lett., 759, L27.
Papadopoulos P. P. et al. (2012a) Mon. Not. R. Astron. Soc., 426,
2601.
Papadopoulos P. P. et al. (2012b) Astrophys. J., 751, 10.
Parker E. N. (1966) Astrophys. J., 145, 811.
Passot T. et al. (1995) Astrophys. J., 455, 536.
Pelupessy F. I. and Papadopoulos P. P. (2009) Astrophys. J., 707,
954.
Pelupessy F. I. et al. (2006) Astrophys. J., 645, 1024.
Penston M. V. et al. (1969) Mon. Not. R. Astron. Soc., 142, 355.
Pety J. et al. (2013) Astrophys. J., 779, 43.
Phillips J. P. (1999) Astron. Astrophys. Suppl., 134, 241.
Pineda J. E. et al. (2009) Astrophys. J. Lett., 699, L134.
Piontek R. A. and Ostriker E. C. (2004) Astrophys. J., 601, 905.
Piontek R. A. and Ostriker E. C. (2005) Astrophys. J., 629, 849.
Piontek R. A. and Ostriker E. C. (2007) Astrophys. J., 663, 183.
Pon A. et al. (2012) Astrophys. J., 756, 145.
Price D. J. et al. (2011) Astrophys. J. Lett., 727, L21.
Rafelski M. et al. (2011) Astrophys. J., 736, 48.
Rahman N. et al. (2012) Astrophys. J., 745, 2, 183.
Rebolledo D. et al. (2012) Astrophys. J., 757, 155.
Regan M. W. et al. (2001) Astrophys. J., 561, 1, 218.
Reggiani M. et al. (2011) Astron. Astrophys., 534, A83.
Rho J. and Petre R. (1998) Astrophys. J. Lett., 503, L167.
Ripple F. et al. (2013) Mon. Not. R. Astron. Soc., 431, 2, 1296.
Roberts W. W. (1969) Astrophys. J., 158, 123.
Robertson B. E. and Kravtsov A. V. (2008) Astrophys. J., 680,
1083.
Robitaille T. P. and Whitney B. A. (2010) Astrophys. J. Lett., 710,
L11.
Rogers H. and Pittard J. M. (2013) Mon. Not. R. Astron. Soc., 431,
1337.
Roman Duval J. et al. (2010) Astrophys. J., 723, 492.
Rosolowsky E. and Blitz L. (2005) Astrophys. J., 623, 2, 826.
Rosolowsky E. and Leroy A. (2006) PASP, 118, 1, 590.
Rosolowsky E. et al. (2003) Astrophys. J., 599, 258.
Rowles J. and Froebrich D. (2009) Mon. Not. R. Astron. Soc., 395,
1640.
Rubio M. et al. (1993) Astron. Astrophys., 271, 9.
Saintonge A. et al. (2011) Mon. Not. R. Astron. Soc., 415, 1, 61.
Saintonge A. et al. (2011a) Mon. Not. R. Astron. Soc., 415, 1, 32.
Sánchez-Salcedo F. J. et al. (2002) Astrophys. J., 577, 768.
Sanders D. et al. (1984) Astrophys. J., 276, 1, 182.
Sandstrom K. M. et al. (2012) Astrophys. J., submitted, arXiv
1212.1208.
24
2511.
Vishniac E. T. (1994) Astrophys. J., 428, 186.
Wada K. et al. (2000) Astrophys. J., 540, 797.
Wada K. et al. (2011) Astrophys. J., 735, 1.
Walch S. K. et al. (2012) Mon. Not. R. Astron. Soc., 427, 625.
Walder R. and Folini D. (2000) Astrophys. Space Sci., 274, 343.
Wang P. et al. (2010) Astrophys. J., 709, 27.
Weinreb S. et al. (1963) Nature, 200, 829.
Whitworth A. (1979) Mon. Not. R. Astron. Soc., 186, 59.
Williams J. P. and McKee C. F. (1997) Astrophys. J., 476, 166.
Williams J. P. et al. (1994) Astrophys. J., 428, 693.
Wilson R. W. et al. (1970) Astrophys. J. Lett., 161, L43.
Wolfe A. M. and Chen H.-W. (2006) Astrophys. J., 652, 981.
Wolfire M. G. et al. (2010) Astrophys. J., 716, 1191.
Wong T. and Blitz L. (2002) Astrophys. J., 569, 157.
Wong T. et al. (2011) Astrophys. J.Supplement, 197, 16.
Wong T. et al. (2013) Astrophys. J. Lett., 777, L4.
Wu J. et al. (2005) Astrophys. J., 635, L173.
Wyder T. K. et al. (2009) Astrophys. J., 696, 1834.
Yonekura Y. et al. (1997) Astrophys. J.Supplement, 110, 21.
Young J. S. et al. (1995) Astrophys. J.Supplement Series, 98, 219.
Zamora-Avilés M. et al. (2012) Astrophys. J., 751, 77.
Zuckerman B. and Evans N. J. (1974) Astrophys. J. Lett., 192,
L149.
This 2-column preprint was prepared with the AAS LATEX macros v5.2.
25